cross_selling_prediction.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "buncit.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "01NIiKm7HZWJ",
        "RXvk9S7hpMIW",
        "zVn7Ai3cHE6f",
        "AAXOzd98pRaI",
        "d19fYXcLHE61",
        "13-OAb8dpPjX",
        "7iiS42FsyeBG",
        "LuipAmGC1aNk",
        "X-8157EF3Gdn",
        "QgJPvRZzdoCh"
      ],
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.6.6"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xQilXLGCHE6T",
        "colab_type": "text"
      },
      "source": [
        "# Cross Selling Analysis & Prediction \n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "RXvk9S7hpMIW"
      },
      "source": [
        "# Import Packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2ZTefVitflPv",
        "jupyter": {
          "outputs_hidden": true
        },
        "outputId": "d7c89f01-6cca-49aa-cd71-b6766b9d2f7d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        }
      },
      "source": [
        "!pip install h2o\n",
        "!pip install boruta"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: h2o in /opt/conda/lib/python3.6/site-packages (3.26.0.10)\r\n",
            "Requirement already satisfied: requests in /opt/conda/lib/python3.6/site-packages (from h2o) (2.22.0)\r\n",
            "Requirement already satisfied: colorama>=0.3.8 in /opt/conda/lib/python3.6/site-packages (from h2o) (0.4.1)\r\n",
            "Requirement already satisfied: future in /opt/conda/lib/python3.6/site-packages (from h2o) (0.18.2)\r\n",
            "Requirement already satisfied: tabulate in /opt/conda/lib/python3.6/site-packages/tabulate-0.8.6-py3.6.egg (from h2o) (0.8.6)\r\n",
            "Requirement already satisfied: idna<2.9,>=2.5 in /opt/conda/lib/python3.6/site-packages (from requests->h2o) (2.8)\r\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /opt/conda/lib/python3.6/site-packages (from requests->h2o) (1.24.2)\r\n",
            "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /opt/conda/lib/python3.6/site-packages (from requests->h2o) (3.0.4)\r\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.6/site-packages (from requests->h2o) (2019.9.11)\r\n",
            "Requirement already satisfied: boruta in /opt/conda/lib/python3.6/site-packages (0.3)\r\n",
            "Requirement already satisfied: scipy>=0.17.0 in /opt/conda/lib/python3.6/site-packages (from boruta) (1.3.3)\r\n",
            "Requirement already satisfied: numpy>=1.10.4 in /opt/conda/lib/python3.6/site-packages (from boruta) (1.17.4)\r\n",
            "Requirement already satisfied: scikit-learn>=0.17.1 in /opt/conda/lib/python3.6/site-packages (from boruta) (0.21.3)\r\n",
            "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.6/site-packages (from scikit-learn>=0.17.1->boruta) (0.14.0)\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "XIfAP69YoquV",
        "outputId": "526246a5-157c-4188-ed47-23316745ec85",
        "colab": {}
      },
      "source": [
        "#Basic\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import math\n",
        "from scipy.stats import skew\n",
        "\n",
        "import ast\n",
        "import matplotlib.pyplot as plt\n",
        "from mpl_toolkits import mplot3d\n",
        "from sklearn.decomposition import PCA\n",
        "from imblearn.over_sampling import SMOTENC\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.cluster import MiniBatchKMeans\n",
        "\n",
        "from boruta import BorutaPy\n",
        "from sklearn.ensemble import RandomForestClassifier\n",
        "\n",
        "import h2o\n",
        "from h2o.automl import H2OAutoML\n",
        "from h2o.estimators import H2OXGBoostEstimator,H2ORandomForestEstimator,H2OGradientBoostingEstimator\n",
        "from h2o.estimators.glm import H2OGeneralizedLinearEstimator\n",
        "from h2o.estimators.stackedensemble import H2OStackedEnsembleEstimator\n",
        "\n",
        "import xgboost as xgb\n",
        "from sklearn.metrics import f1_score,confusion_matrix\n",
        "from sklearn.model_selection import KFold,GridSearchCV,train_test_split\n",
        "from sklearn.tree import DecisionTreeClassifier\n",
        "\n",
        "from keras.layers import Input, Dense, Reshape, Embedding\n",
        "from keras.models import Model\n",
        "from keras import backend as K\n",
        "from numpy import array\n",
        "from keras.preprocessing.text import one_hot\n",
        "from keras.preprocessing.sequence import pad_sequences"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "AAXOzd98pRaI"
      },
      "source": [
        "# Import Data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4gFTzQETJK2Z",
        "colab_type": "text"
      },
      "source": [
        "Data used in this modelling process is the flight (CGK-DPS) + hotel transaction history on 2018 from one of the biggest ticketing company in Indonesia "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "mFitVGPWyMSS",
        "jupyter": {
          "outputs_hidden": true
        },
        "colab": {}
      },
      "source": [
        "df_hotel = pd.read_csv('/kaggle/input/datavidia2019/hotel.csv')\n",
        "df = pd.read_csv('/kaggle/input/datavidia2019/flight.csv')\n",
        "df_test=pd.read_csv('/kaggle/input/datavidia2019/test.csv')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sWpM2iczHE6p",
        "colab_type": "text"
      },
      "source": [
        "The target variable in this modelling process is **is_cross_sell** which has value \"Yes\" when the  **hotel_id** is not *None* and has value \"Np\" when **hotel_id** is  *None*"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sAXzakPRHE6q",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df['is_cross_sell']=df['hotel_id'].apply(lambda x: 'Yes' if x !='None' else 'No')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3QCacySIHE6u",
        "colab_type": "text"
      },
      "source": [
        "Below is the snippet of flight.csv and hotel.csv data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "zThlCO74t8Jy",
        "outputId": "339aa630-48a3-44fc-8fdb-84095e5eb88e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 484
        }
      },
      "source": [
        "df.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>account_id</th>\n",
              "      <th>order_id</th>\n",
              "      <th>member_duration_days</th>\n",
              "      <th>gender</th>\n",
              "      <th>trip</th>\n",
              "      <th>service_class</th>\n",
              "      <th>price</th>\n",
              "      <th>is_tx_promo</th>\n",
              "      <th>no_of_seats</th>\n",
              "      <th>airlines_name</th>\n",
              "      <th>route</th>\n",
              "      <th>hotel_id</th>\n",
              "      <th>visited_city</th>\n",
              "      <th>log_transaction</th>\n",
              "      <th>is_cross_sell</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>912aa410a02cd7e1bab414214a7005c0</td>\n",
              "      <td>5c6f39c690f23650d3cde28e5b51c908</td>\n",
              "      <td>566.0</td>\n",
              "      <td>M</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>885898.00</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>33199710eb822fbcfd0dc793f4788d30</td>\n",
              "      <td>CGK - DPS</td>\n",
              "      <td>None</td>\n",
              "      <td>'['Semarang', 'Jakarta', 'Medan', 'Bali']'</td>\n",
              "      <td>'[545203.03, 918492.11, 1774241.4, 885898.0]'</td>\n",
              "      <td>No</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>d64a90a618202a5e8b25d8539377f3ca</td>\n",
              "      <td>5cbef2b87f51c18bf399d11bfe495a46</td>\n",
              "      <td>607.0</td>\n",
              "      <td>M</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>2139751.25</td>\n",
              "      <td>NO</td>\n",
              "      <td>2.0</td>\n",
              "      <td>0a102015e48c1f68e121acc99fca9a05</td>\n",
              "      <td>CGK - DPS</td>\n",
              "      <td>None</td>\n",
              "      <td>'['Jakarta', 'Medan', 'Bali']'</td>\n",
              "      <td>'[555476.36, 2422826.84, 7398697.64, 7930866.8...</td>\n",
              "      <td>No</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1a42ac02bcb4a902973123323f84da55</td>\n",
              "      <td>38fc35a1e62384012a358ab1fbd5ad03</td>\n",
              "      <td>648.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>2695550.00</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0a102015e48c1f68e121acc99fca9a05</td>\n",
              "      <td>CGK - DPS</td>\n",
              "      <td>None</td>\n",
              "      <td>'['Semarang', 'Jakarta', 'Medan', 'Bali']'</td>\n",
              "      <td>'[7328957.45, 7027662.34, 1933360.88, 3461836....</td>\n",
              "      <td>No</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>92cddd64d4be4dec6dfbcc0c50e902f4</td>\n",
              "      <td>c7f54cb748828b4413e02dea2758faf6</td>\n",
              "      <td>418.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>1146665.00</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0a102015e48c1f68e121acc99fca9a05</td>\n",
              "      <td>CGK - DPS</td>\n",
              "      <td>None</td>\n",
              "      <td>'['Jogjakarta', 'Bali', 'Jakarta', 'Medan']'</td>\n",
              "      <td>'[5243631.69, 2474344.48, 1146665.0]'</td>\n",
              "      <td>No</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>bf637abc47ea93bad22264f4956d67f6</td>\n",
              "      <td>dec228e4d2b6023c9f1fe9cfe9c451bf</td>\n",
              "      <td>537.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>1131032.50</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>6c483c0812c96f8ec43bb0ff76eaf716</td>\n",
              "      <td>CGK - DPS</td>\n",
              "      <td>None</td>\n",
              "      <td>'['Jakarta', 'Bali', 'Medan', 'Jogjakarta', 'S...</td>\n",
              "      <td>'[9808972.98, 9628619.79, 6712680.0, 5034510.0...</td>\n",
              "      <td>No</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                         account_id                          order_id  \\\n",
              "0  912aa410a02cd7e1bab414214a7005c0  5c6f39c690f23650d3cde28e5b51c908   \n",
              "1  d64a90a618202a5e8b25d8539377f3ca  5cbef2b87f51c18bf399d11bfe495a46   \n",
              "2  1a42ac02bcb4a902973123323f84da55  38fc35a1e62384012a358ab1fbd5ad03   \n",
              "3  92cddd64d4be4dec6dfbcc0c50e902f4  c7f54cb748828b4413e02dea2758faf6   \n",
              "4  bf637abc47ea93bad22264f4956d67f6  dec228e4d2b6023c9f1fe9cfe9c451bf   \n",
              "\n",
              "   member_duration_days gender  trip service_class       price is_tx_promo  \\\n",
              "0                 566.0      M  trip       ECONOMY   885898.00          NO   \n",
              "1                 607.0      M  trip       ECONOMY  2139751.25          NO   \n",
              "2                 648.0      F  trip       ECONOMY  2695550.00          NO   \n",
              "3                 418.0      F  trip       ECONOMY  1146665.00          NO   \n",
              "4                 537.0      F  trip       ECONOMY  1131032.50          NO   \n",
              "\n",
              "   no_of_seats                     airlines_name      route hotel_id  \\\n",
              "0          1.0  33199710eb822fbcfd0dc793f4788d30  CGK - DPS     None   \n",
              "1          2.0  0a102015e48c1f68e121acc99fca9a05  CGK - DPS     None   \n",
              "2          1.0  0a102015e48c1f68e121acc99fca9a05  CGK - DPS     None   \n",
              "3          1.0  0a102015e48c1f68e121acc99fca9a05  CGK - DPS     None   \n",
              "4          1.0  6c483c0812c96f8ec43bb0ff76eaf716  CGK - DPS     None   \n",
              "\n",
              "                                        visited_city  \\\n",
              "0         '['Semarang', 'Jakarta', 'Medan', 'Bali']'   \n",
              "1                     '['Jakarta', 'Medan', 'Bali']'   \n",
              "2         '['Semarang', 'Jakarta', 'Medan', 'Bali']'   \n",
              "3       '['Jogjakarta', 'Bali', 'Jakarta', 'Medan']'   \n",
              "4  '['Jakarta', 'Bali', 'Medan', 'Jogjakarta', 'S...   \n",
              "\n",
              "                                     log_transaction is_cross_sell  \n",
              "0      '[545203.03, 918492.11, 1774241.4, 885898.0]'            No  \n",
              "1  '[555476.36, 2422826.84, 7398697.64, 7930866.8...            No  \n",
              "2  '[7328957.45, 7027662.34, 1933360.88, 3461836....            No  \n",
              "3              '[5243631.69, 2474344.48, 1146665.0]'            No  \n",
              "4  '[9808972.98, 9628619.79, 6712680.0, 5034510.0...            No  "
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AMYIQWp_HE6z",
        "colab_type": "code",
        "outputId": "f708ab05-84a7-4901-b420-6f6af9311b0f",
        "colab": {}
      },
      "source": [
        "df_hotel.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hotel_id</th>\n",
              "      <th>starRating</th>\n",
              "      <th>city</th>\n",
              "      <th>free_wifi</th>\n",
              "      <th>pool_access</th>\n",
              "      <th>free_breakfast</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>e2733e84102226acf6b53bffd2e60cf8</td>\n",
              "      <td>0.0</td>\n",
              "      <td>bali</td>\n",
              "      <td>YES</td>\n",
              "      <td>NO</td>\n",
              "      <td>NO</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>9f9de5df06d64ada1026e930687a87e4</td>\n",
              "      <td>0.0</td>\n",
              "      <td>bali</td>\n",
              "      <td>YES</td>\n",
              "      <td>NO</td>\n",
              "      <td>NO</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3cf6774fb4dc331bb49e7a959b74a67e</td>\n",
              "      <td>0.0</td>\n",
              "      <td>bali</td>\n",
              "      <td>YES</td>\n",
              "      <td>NO</td>\n",
              "      <td>NO</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>eca261898220478834072b0c753a5229</td>\n",
              "      <td>0.0</td>\n",
              "      <td>bali</td>\n",
              "      <td>YES</td>\n",
              "      <td>NO</td>\n",
              "      <td>NO</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>c21f400013fa4f244a7168a3c155b8b5</td>\n",
              "      <td>0.0</td>\n",
              "      <td>bali</td>\n",
              "      <td>YES</td>\n",
              "      <td>NO</td>\n",
              "      <td>NO</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                           hotel_id  starRating  city free_wifi pool_access  \\\n",
              "0  e2733e84102226acf6b53bffd2e60cf8         0.0  bali       YES          NO   \n",
              "1  9f9de5df06d64ada1026e930687a87e4         0.0  bali       YES          NO   \n",
              "2  3cf6774fb4dc331bb49e7a959b74a67e         0.0  bali       YES          NO   \n",
              "3  eca261898220478834072b0c753a5229         0.0  bali       YES          NO   \n",
              "4  c21f400013fa4f244a7168a3c155b8b5         0.0  bali       YES          NO   \n",
              "\n",
              "  free_breakfast  \n",
              "0             NO  \n",
              "1             NO  \n",
              "2             NO  \n",
              "3             NO  \n",
              "4             NO  "
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d19fYXcLHE61",
        "colab_type": "text"
      },
      "source": [
        "# Exploratory data analysis "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "Here we defined some functions used for EDA"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "def box_hist_plot(df,column):\n",
        "  '''\n",
        "  Function for boxplot and histogram in 1 figure\n",
        "  '''\n",
        "  plt.figure(figsize=(15,6))\n",
        "  plt.subplot(1, 2, 1)\n",
        "  df[[column]].boxplot()\n",
        "\n",
        "  plt.subplot(1, 2, 2)\n",
        "  df[column].hist()\n",
        "  \n",
        "\n",
        "def detect_outlier(df,column,thres=1.5,plot=False):\n",
        "  '''\n",
        "  Function to detect outlier\n",
        "  '''\n",
        "  IQR = df[column].quantile(0.75) - df[column].quantile(0.25)\n",
        "  upper_fence = df[column].quantile(0.75) + (IQR * thres)\n",
        "  lower_fence = df[column].quantile(0.25)  - (IQR * thres)\n",
        "  \n",
        "  print('Upper Fence: {}\\n'.format(upper_fence))\n",
        "  print('Lower Fence: {}\\n'.format(lower_fence))\n",
        "  print('Proportion of Upper Outliers: {}\\n'.format(len(df[df[column]>upper_fence])/len(df)*100))\n",
        "  print('Proportion of Lower Outliers: {}\\n'.format(len(df[df[column]<lower_fence])/len(df)*100))\n",
        "  \n",
        "  if plot==True:\n",
        "    box_hist_plot(df,column)\n",
        "\n",
        "\n",
        "def correlated(df,tolerance):\n",
        "  '''\n",
        "  Function to know correlated variables\n",
        "  Input: {dataframe, correlation_tolerance}\n",
        "  '''\n",
        "  cate_features_index = np.where(df.dtypes == 'O')[0]\n",
        "  num_features_index = [x for x in range(len(df.columns)) if x not in cate_features_index]\n",
        "  \n",
        "  #Check correlation of numerical variables\n",
        "  corrmat = df.iloc[:,num_features_index].corr()\n",
        "  corrmat = corrmat.abs().unstack() # absolute value of corr coef\n",
        "  corrmat = corrmat.sort_values(ascending=False)\n",
        "  corrmat = corrmat[corrmat >= tolerance]\n",
        "  corrmat = corrmat[corrmat < 1]\n",
        "  corrmat = pd.DataFrame(corrmat).reset_index()\n",
        "  corrmat.columns = ['feature1', 'feature2', 'corr']\n",
        "\n",
        "  # find groups of correlated features\n",
        "\n",
        "  grouped_feature_ls = []\n",
        "  correlated_groups = []\n",
        "\n",
        "  for feature in corrmat.feature1.unique():\n",
        "    if feature not in grouped_feature_ls:\n",
        "      # find all features correlated to a single feature\n",
        "      correlated_block = corrmat[corrmat.feature1 == feature]\n",
        "      grouped_feature_ls = grouped_feature_ls + list(\n",
        "      correlated_block.feature2.unique()) + [feature]\n",
        "\n",
        "      # append the block of features to the list\n",
        "      correlated_groups.append(correlated_block)\n",
        "\n",
        "  print('found {} correlated groups'.format(len(correlated_groups)))\n",
        "  print('out of {} total features'.format(df.iloc[:,num_features_index].shape[1]))\n",
        "\n",
        "  for group in correlated_groups:\n",
        "    print(group)\n",
        "    print()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "13-OAb8dpPjX"
      },
      "source": [
        "## Flight Data "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FxeGP5huKKdA",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "Below is the flight.csv data info"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7MVuL-jkKGJH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df.info()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1HVkKfHxHE62",
        "colab_type": "text"
      },
      "source": [
        "Below is the frequency table from ordered  **hotel_id**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "8n7HXZ8xgUHN",
        "outputId": "cbbf9c1a-9dad-42f0-b2b1-e08f7ea49201",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 634
        }
      },
      "source": [
        "pd.DataFrame(df['hotel_id'].value_counts()).head(20)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>hotel_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>None</th>\n",
              "      <td>111198</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cd8210b5c97be3564a4cd0283dda323f</th>\n",
              "      <td>178</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8631d3e2f6a000ddbfcef3a4a547d642</th>\n",
              "      <td>107</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cb1838f895d879993c8ca9eaf729dbeb</th>\n",
              "      <td>94</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>a3bc533c2761aa779605f68011ee8688</th>\n",
              "      <td>90</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>e366b55a2e316618559ef5dd0b4a534f</th>\n",
              "      <td>87</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50a5b6704042f4abde9e6c365cc5ec24</th>\n",
              "      <td>80</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>d98370364c38275cf0176b3c037caeb3</th>\n",
              "      <td>78</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4f640616d3102097c41affeadb05c606</th>\n",
              "      <td>76</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>79f4f7137c68a6be56d8b161114e7adb</th>\n",
              "      <td>75</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>d4a022fbcb4e07740c08c6aa079abdcc</th>\n",
              "      <td>73</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>d4dee0b2c32b586b2438d4af18d8606e</th>\n",
              "      <td>69</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>d160957744810bce2098bb7ec1097102</th>\n",
              "      <td>68</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ae7e0a7b8e8133fb5f089a86697c4afd</th>\n",
              "      <td>62</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fa736b7ceaa1f2eb5dcf02532fa61368</th>\n",
              "      <td>61</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ad223e69b076f38d0e08f7892f64cad8</th>\n",
              "      <td>57</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>c329d21f9735f1bca222eac7ef6a9a23</th>\n",
              "      <td>55</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ec31f16cbed9dc70ba8c1f189c6eb9d5</th>\n",
              "      <td>55</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6ef62d2cb35181a07dd9be08f6cf6d6c</th>\n",
              "      <td>54</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>96162de076749853406edfc328d8efad</th>\n",
              "      <td>53</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                  hotel_id\n",
              "None                                111198\n",
              "cd8210b5c97be3564a4cd0283dda323f       178\n",
              "8631d3e2f6a000ddbfcef3a4a547d642       107\n",
              "cb1838f895d879993c8ca9eaf729dbeb        94\n",
              "a3bc533c2761aa779605f68011ee8688        90\n",
              "e366b55a2e316618559ef5dd0b4a534f        87\n",
              "50a5b6704042f4abde9e6c365cc5ec24        80\n",
              "d98370364c38275cf0176b3c037caeb3        78\n",
              "4f640616d3102097c41affeadb05c606        76\n",
              "79f4f7137c68a6be56d8b161114e7adb        75\n",
              "d4a022fbcb4e07740c08c6aa079abdcc        73\n",
              "d4dee0b2c32b586b2438d4af18d8606e        69\n",
              "d160957744810bce2098bb7ec1097102        68\n",
              "ae7e0a7b8e8133fb5f089a86697c4afd        62\n",
              "fa736b7ceaa1f2eb5dcf02532fa61368        61\n",
              "ad223e69b076f38d0e08f7892f64cad8        57\n",
              "c329d21f9735f1bca222eac7ef6a9a23        55\n",
              "ec31f16cbed9dc70ba8c1f189c6eb9d5        55\n",
              "6ef62d2cb35181a07dd9be08f6cf6d6c        54\n",
              "96162de076749853406edfc328d8efad        53"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2hJtBNWcHE65",
        "colab_type": "text"
      },
      "source": [
        "111198 orders have *None* values which means those orders are not classified as cross selling order or it can be said that only around 5% of orders which are classified as cross selling orders, means that this is a *imbalanced* classification problem"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iQ-FdhYOHE66",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**account_id**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r6liQqPFHE67",
        "colab_type": "code",
        "outputId": "6f3e25a8-a442-42bc-ffff-aca797851cbc",
        "colab": {}
      },
      "source": [
        "pd.DataFrame(df['account_id'].value_counts()).head(20)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>account_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>d64a90a618202a5e8b25d8539377f3ca</th>\n",
              "      <td>3311</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fbb4123bb341ad85719af86930a0ccea</th>\n",
              "      <td>311</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>bf637abc47ea93bad22264f4956d67f6</th>\n",
              "      <td>161</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>74b423f5354172076e9529debf8866ac</th>\n",
              "      <td>145</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4944140d10e202e845f0766a3e3ac2d1</th>\n",
              "      <td>133</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>fabbb09276eea6bcf6cc7baaed9d716a</th>\n",
              "      <td>118</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>49aef087188299dca09278b382147b44</th>\n",
              "      <td>117</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ea9b9a3dc459101446cc47ad48d336c0</th>\n",
              "      <td>106</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>c3d5f862188aaa1299d0cc99142b4981</th>\n",
              "      <td>105</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5634d6fd8d25ce0fd43b00a26d6dc17c</th>\n",
              "      <td>105</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>f7b64bbf8462b0d1a006af1a983c609d</th>\n",
              "      <td>104</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>721b125ee977ecb81560cf05a50aae9f</th>\n",
              "      <td>93</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>965149909f4e9e332a9dc1c5c9b3ad99</th>\n",
              "      <td>70</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>b9bab0e84361493561c68ca49f024bf9</th>\n",
              "      <td>68</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>e5246f41c61f6fde9429e9255a257c01</th>\n",
              "      <td>67</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6672682f0d5881bb7a49e8b8af28efeb</th>\n",
              "      <td>63</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>f96e824a7177206a22adbc7c8e6fd29a</th>\n",
              "      <td>62</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>b4f1bd73bea184d2f6183fcc327a68de</th>\n",
              "      <td>62</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>237a26e0c757dde8225b8a204ea5be65</th>\n",
              "      <td>61</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10cbe5d3abd5451406703bf00d0a0af3</th>\n",
              "      <td>57</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                  account_id\n",
              "d64a90a618202a5e8b25d8539377f3ca        3311\n",
              "fbb4123bb341ad85719af86930a0ccea         311\n",
              "bf637abc47ea93bad22264f4956d67f6         161\n",
              "74b423f5354172076e9529debf8866ac         145\n",
              "4944140d10e202e845f0766a3e3ac2d1         133\n",
              "fabbb09276eea6bcf6cc7baaed9d716a         118\n",
              "49aef087188299dca09278b382147b44         117\n",
              "ea9b9a3dc459101446cc47ad48d336c0         106\n",
              "c3d5f862188aaa1299d0cc99142b4981         105\n",
              "5634d6fd8d25ce0fd43b00a26d6dc17c         105\n",
              "f7b64bbf8462b0d1a006af1a983c609d         104\n",
              "721b125ee977ecb81560cf05a50aae9f          93\n",
              "965149909f4e9e332a9dc1c5c9b3ad99          70\n",
              "b9bab0e84361493561c68ca49f024bf9          68\n",
              "e5246f41c61f6fde9429e9255a257c01          67\n",
              "6672682f0d5881bb7a49e8b8af28efeb          63\n",
              "f96e824a7177206a22adbc7c8e6fd29a          62\n",
              "b4f1bd73bea184d2f6183fcc327a68de          62\n",
              "237a26e0c757dde8225b8a204ea5be65          61\n",
              "10cbe5d3abd5451406703bf00d0a0af3          57"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EpKQL1qjHE6-",
        "colab_type": "text"
      },
      "source": [
        "There is a user which has more than 3000 transactions in 1 year. This account will be classified as the outliers and will be removed from the data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Ig4CPjios-97",
        "colab": {}
      },
      "source": [
        "print('Unique numbers of acount_id : {}'.format(len(df['account_id'].unique())))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Y47b2DPEtuSa",
        "outputId": "97c5e4be-b23a-4f37-8d01-bdfa7cae1ed1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "print('Unique numbers of order_id: {}'.format(len(df['order_id'].unique())))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Banyaknya order_id yang unik: 117946\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JbgOxwkFHE7J",
        "colab_type": "text"
      },
      "source": [
        "Number of unique **order_id** is same as the number of rows in the data which means there is no duplicated order in the data\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jSqqrPDZHE7L",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**member_duration_days**\n",
        "\n",
        "Below is the descriptive statistics of **member_duration_days**  variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "Axo9FCgmujzx",
        "outputId": "8de5f090-3de9-4666-c1a9-7058040018e2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        }
      },
      "source": [
        "df[['member_duration_days']].describe()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>member_duration_days</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>117946.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>955.586955</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>565.292428</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>329.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>537.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>678.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>1341.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>2720.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "       member_duration_days\n",
              "count         117946.000000\n",
              "mean             955.586955\n",
              "std              565.292428\n",
              "min              329.000000\n",
              "25%              537.000000\n",
              "50%              678.000000\n",
              "75%             1341.000000\n",
              "max             2720.000000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0IWhLvq5HE7S",
        "colab_type": "text"
      },
      "source": [
        "The average member duration for user which has buy a flight tickets is around 2.6 years\n",
        "\n",
        "It will be check if there is outlier for this variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "sZe6Ft8WwrWv",
        "outputId": "15db8727-0f14-4b6f-afe3-a8488df6772d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 510
        }
      },
      "source": [
        "detect_outlier(df,'member_duration_days',plot=True)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Upper Fence: 2547.0\n",
            "\n",
            "Lower Fence: -669.0\n",
            "\n",
            "Proportion of Upper Outliers: 0.2916588947484442\n",
            "\n",
            "Proportion of Lower Outliers: 0.0\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l4xEeUAhHE7b",
        "colab_type": "text"
      },
      "source": [
        "From the data plot, **member_duration_days** variable has the skew to the right tendency, resulting some outliers in the high value of this variable (above the upper whisker value)\n",
        "\n",
        "It will be checked the cross selling behaviour from the outlier users (**member_duration_days** above 2547 days)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "9hECr8e34GL5",
        "outputId": "ac6cafae-e232-4eb2-9a50-942dea2a539b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df[df['member_duration_days']>2547]['is_cross_sell'].value_counts()/len(df[df['member_duration_days']>2547])*100"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "No     93.895349\n",
              "Yes     6.104651\n",
              "Name: is_cross_sell, dtype: float64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "gIBkhdmc4Yo7",
        "outputId": "5715f1ee-466d-4a9f-9068-f3344236c797",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df[df['member_duration_days']<=2547]['is_cross_sell'].value_counts()/len(df[df['member_duration_days']<=2547])*100"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "No     94.279859\n",
              "Yes     5.720141\n",
              "Name: is_cross_sell, dtype: float64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "99gZTg4iHE7j",
        "colab_type": "text"
      },
      "source": [
        "It can be seen that the outlier user still has the similar behaviour with non outlier users"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_7xVf8rrHE7m",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**price**\n",
        "\n",
        "Below is the descriptive statistics for **price** variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "HA6CamqWutlk",
        "outputId": "c597f5a0-a7c8-4a24-fd3a-dd7f04a7ce2c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        }
      },
      "source": [
        "df[['price']].describe()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>price</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>1.179460e+05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>2.672772e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>2.449247e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>4.069520e+05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>1.282710e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>1.987061e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>2.974400e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>5.903643e+07</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "              price\n",
              "count  1.179460e+05\n",
              "mean   2.672772e+06\n",
              "std    2.449247e+06\n",
              "min    4.069520e+05\n",
              "25%    1.282710e+06\n",
              "50%    1.987061e+06\n",
              "75%    2.974400e+06\n",
              "max    5.903643e+07"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m4KD7CkNHE7r",
        "colab_type": "text"
      },
      "source": [
        "The average value of price from the user which bough flight ticket is around IDR 2.6 million / transactions\n",
        "\n",
        "It will be checked if there is an outlier for this variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "VeSeSfERwteo",
        "outputId": "b6c6243e-9fbc-409b-a974-59b474cb86d8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 533
        }
      },
      "source": [
        "detect_outlier(df,'price',plot=True)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Upper Fence: 5511935.0\n",
            "\n",
            "Lower Fence: -1254825.0\n",
            "\n",
            "Proportion of Upper Outliers: 8.324148339070423\n",
            "\n",
            "Proportion of Lower Outliers: 0.0\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x432 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jdFyJP4DHE7u",
        "colab_type": "text"
      },
      "source": [
        "**price** variable has a more positive skewness compared to **member_duration_days**\n",
        "\n",
        "Same with **member_duration_days**, it will be cheked the cross selling behaviour of the outlier users."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "tQCZchfw5X8w",
        "outputId": "2b113ce4-8c2a-43c7-ea42-e5b18720952f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df[df['price']>5511935]['is_cross_sell'].value_counts()/len(df[df['price']>5511935])*100"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "No     93.786922\n",
              "Yes     6.213078\n",
              "Name: is_cross_sell, dtype: float64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "4jWl7wRX5etp",
        "outputId": "8d843468-defa-4356-d01a-bdee0df63d02",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df[df['price']<=5511935]['is_cross_sell'].value_counts()/len(df[df['price']<=5511935])*100"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "No     94.323394\n",
              "Yes     5.676606\n",
              "Name: is_cross_sell, dtype: float64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bA9d8m4CHE73",
        "colab_type": "text"
      },
      "source": [
        "It can be seen that the outlier user still has the similar behaviour with non outlier users"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5c9lr2s0HE73",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**no_of_seats**\n",
        "\n",
        "Below is the frequency table for **no_of_seats**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "PYN48KFMUjLq",
        "outputId": "ab868220-a58d-4230-af56-3c09dc1cb106",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 220
        }
      },
      "source": [
        "df['no_of_seats'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1.0     70937\n",
              "2.0     29561\n",
              "3.0      8236\n",
              "4.0      4963\n",
              "5.0      2092\n",
              "6.0      1165\n",
              "7.0       882\n",
              "8.0        75\n",
              "9.0        31\n",
              "10.0        3\n",
              "11.0        1\n",
              "Name: no_of_seats, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uMpzt8fJHE75",
        "colab_type": "text"
      },
      "source": [
        "We have a hypothesis that the high number of orders caused by the high number of **no_of_seats**\n",
        "\n",
        "Below is the frequency table of **no_of_seats** from the orders **price** above 5511935"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2ejMGI-cTqJj",
        "outputId": "4239f5a6-fe17-4799-e0ee-5dbc8be020c4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 220
        }
      },
      "source": [
        "df[df['price']>5511935]['no_of_seats'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "4.0     2180\n",
              "2.0     1933\n",
              "3.0     1855\n",
              "5.0     1369\n",
              "6.0     1037\n",
              "7.0      861\n",
              "1.0      475\n",
              "8.0       73\n",
              "9.0       31\n",
              "10.0       3\n",
              "11.0       1\n",
              "Name: no_of_seats, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5JJ55_BaHE78",
        "colab_type": "text"
      },
      "source": [
        "High valued transactions are dominated by the orders which has the high number of **no_of_seats** "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "eO08aTYVxHZM",
        "jupyter": {
          "outputs_hidden": true
        },
        "outputId": "2898d84a-e35f-4616-8034-beaf975afe3e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 101
        }
      },
      "source": [
        "# Checking if there is a pair of variable which are correlated \n",
        "correlated(df,0.7)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "found 1 correlated groups\n",
            "out of 3 total features\n",
            "      feature1 feature2      corr\n",
            "0  no_of_seats    price  0.712863\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pUGX9THVHE7-",
        "colab_type": "text"
      },
      "source": [
        "Only **price** and **no_of_seats** has a pretty strong correlation value"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "juxEFAtPHE7_",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**service_class**\n",
        "\n",
        "Below is the frequency table for **service_class** variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fJdGu4MNHE8B",
        "colab_type": "code",
        "outputId": "2fd3d85f-9416-49c9-ed3b-b660ce158100",
        "colab": {}
      },
      "source": [
        "df['service_class'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "ECONOMY     117665\n",
              "BUSINESS       281\n",
              "Name: service_class, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-n32-0ytHE8F",
        "colab_type": "text"
      },
      "source": [
        "Only around 0.2% of the transactions is the business class flight"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9ZqclgSIHE8F",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**gender**\n",
        "\n",
        "Below is the frequency table for **gender**  variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5QoNh1GEHE8G",
        "colab_type": "code",
        "outputId": "ce6cff00-01ab-46f0-86d5-5d95021707f8",
        "colab": {}
      },
      "source": [
        "df['gender'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "M       61036\n",
              "F       56886\n",
              "None       24\n",
              "Name: gender, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P0r7y-iDHE8I",
        "colab_type": "text"
      },
      "source": [
        "Number of transactions transacted by man is higher than transacted by woman. \n",
        "\n",
        "There are 24 **Missing Values** for the **gender** variable"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "syxq77wOHE8I",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**is_tx_promo**\n",
        "\n",
        "Below is the frequency table for  **is_tx_promo** variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WlGV6Tf2HE8J",
        "colab_type": "code",
        "outputId": "7e7275b7-8e2d-44f6-8dee-05d31bafae3e",
        "colab": {}
      },
      "source": [
        "df['is_tx_promo'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NO     67621\n",
              "YES    50325\n",
              "Name: is_tx_promo, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_XrYRbHuHE8L",
        "colab_type": "text"
      },
      "source": [
        "Interesting! Order which use promo is pretty enough, almost have 1:1 amount ratio with non promo order"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "7iiS42FsyeBG"
      },
      "source": [
        "---\n",
        "\n",
        "## Hotel Data  "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XO9-BJPAHE8M",
        "colab_type": "text"
      },
      "source": [
        "Below is the descriptive statistics for **starRating** variable"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "ozNgJXisyffz",
        "outputId": "e3f4c46d-f1d6-4803-981f-188f16ff036c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        }
      },
      "source": [
        "df_hotel.describe()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>starRating</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>2962.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>2.992235</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>1.270136</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>2.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>3.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>4.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>5.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "        starRating\n",
              "count  2962.000000\n",
              "mean      2.992235\n",
              "std       1.270136\n",
              "min       0.000000\n",
              "25%       2.000000\n",
              "50%       3.000000\n",
              "75%       4.000000\n",
              "max       5.000000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TdoppBl7HE8Q",
        "colab_type": "text"
      },
      "source": [
        "In average, hotel in Bali is 3 star hotel. \n",
        "\n",
        "There are 2962 hotels in the data. However, when we check the number of hotels in Bali, there are more than 10000 hotels. This means :\n",
        "1. Number of hotel in 2019 is way more than 2018; or\n",
        "2. This hotel data is not complete; or\n",
        "3. This data only covered the hotel partner with the company"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "v4TZMtCdPP6B",
        "outputId": "65bb9284-9e2f-407d-eff0-d70589b456e6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 135
        }
      },
      "source": [
        "df.groupby('is_cross_sell').median()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>member_duration_days</th>\n",
              "      <th>price</th>\n",
              "      <th>no_of_seats</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>is_cross_sell</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>No</th>\n",
              "      <td>675.0</td>\n",
              "      <td>1984060.0</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Yes</th>\n",
              "      <td>831.0</td>\n",
              "      <td>2169960.0</td>\n",
              "      <td>1.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               member_duration_days      price  no_of_seats\n",
              "is_cross_sell                                              \n",
              "No                            675.0  1984060.0          1.0\n",
              "Yes                           831.0  2169960.0          1.0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "LuipAmGC1aNk"
      },
      "source": [
        "---\n",
        "## Join Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "gV8owe3o1cWK",
        "colab": {}
      },
      "source": [
        "df_join=df.merge(df_hotel,on='hotel_id',how='left')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "To8awi7Y2a4t",
        "outputId": "759dbe3d-8096-48f5-e548-720b858a4f4b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "unique_hotel_id=df_join[df_join.is_cross_sell=='Yes']['hotel_id'].unique()\n",
        "print('Number of unique hotels transacted by customer: {}'.format(len(unique_hotel_id)))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of unique hotels transacted by customer: 675\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "A6b1b3Pe7Kqr",
        "colab": {}
      },
      "source": [
        "df_hotel.index=df_hotel.hotel_id"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "x76Gq0CD8kiW",
        "outputId": "0e10855f-0e70-479c-aa55-f286b9391f68",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df_hotel.loc[unique_hotel_id]['free_breakfast'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NO     413\n",
              "YES    182\n",
              "Name: free_breakfast, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 31
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "6osiGOGh99VD",
        "outputId": "7a8d1451-0d63-45ec-b9e2-36fb1d656751",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df_hotel.loc[unique_hotel_id]['free_wifi'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "YES    593\n",
              "NO       2\n",
              "Name: free_wifi, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "aVe3edvy_YxI",
        "outputId": "4c9672d5-b5fe-4320-db0e-cf3daa142ee8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        }
      },
      "source": [
        "df_hotel.loc[unique_hotel_id]['pool_access'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "NO     592\n",
              "YES      3\n",
              "Name: pool_access, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "8gloXFED_q9u",
        "jupyter": {
          "outputs_hidden": true
        },
        "outputId": "9589e728-58c7-47fc-8353-8b9cdfe94a6c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 135
        }
      },
      "source": [
        "df_hotel.loc[unique_hotel_id]['starRating'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "3.0    221\n",
              "4.0    197\n",
              "5.0     84\n",
              "2.0     66\n",
              "1.0     24\n",
              "0.0      3\n",
              "Name: starRating, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "b1oH7hjmHE8w",
        "colab_type": "text"
      },
      "source": [
        "The most ordered hotel is 3 star hotel"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "X-8157EF3Gdn"
      },
      "source": [
        "# Data Cleaning/Pre-processing "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "emYMSyFOHE8z",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "*Missing values* on **gender**\n",
        "\n",
        "Because there are only 24 missing values, it will be imputed by the most dominant **gender** values which is *M*"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ggVhr-uGHE80",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df['gender']=df['gender'].apply(lambda x: 'M' if x=='None' else x)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "FUM_ICZeJvdl",
        "outputId": "af015d84-a821-4d20-a4a9-37c3cac5b6bf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 169
        }
      },
      "source": [
        "df['airlines_name'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "6c483c0812c96f8ec43bb0ff76eaf716    36669\n",
              "33199710eb822fbcfd0dc793f4788d30    30640\n",
              "0a102015e48c1f68e121acc99fca9a05    20928\n",
              "ad5bef60d81ea077018f4d50b813153a    11890\n",
              "74c5549aa99d55280a896ea50068a211    10751\n",
              "e35de6a36d385711a660c72c0286154a     7060\n",
              "9855a1d3de1c46526dde37c5d6fb758c        6\n",
              "6872b49542519aea7ae146e23fab5c08        2\n",
              "Name: airlines_name, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "d2nKxa2sg45_",
        "outputId": "5dccc1f6-5fe3-4e55-d6f7-85311b90eb37",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 169
        }
      },
      "source": [
        "df['airlines_name'].value_counts()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "6c483c0812c96f8ec43bb0ff76eaf716    36669\n",
              "33199710eb822fbcfd0dc793f4788d30    30640\n",
              "0a102015e48c1f68e121acc99fca9a05    20928\n",
              "ad5bef60d81ea077018f4d50b813153a    11890\n",
              "74c5549aa99d55280a896ea50068a211    10751\n",
              "e35de6a36d385711a660c72c0286154a     7060\n",
              "9855a1d3de1c46526dde37c5d6fb758c        6\n",
              "6872b49542519aea7ae146e23fab5c08        2\n",
              "Name: airlines_name, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 37
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0E0_22OlHE88",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "\n",
        "\n",
        "Inconsistent **trip** values\n",
        "\n",
        "Trip is an categorical variable which has 3 unique values:\n",
        "\n",
        "*   trip\n",
        "*   round\n",
        "* roundtrip\n",
        "\n",
        "It is a common sense that *round* and *roundtrip* has a same meaning, so the *roundtrip* value is converted into *round*"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N79_GcQpHE8-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df['trip']=df['trip'].apply(lambda x: 'round' if x=='roundtrip' else x)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bsLrFgufHE9A",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**airlines_name** uncommon names\n",
        "\n",
        "In order to ease the identification process, the airlines name will be converted into 1 char code"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pvQnKTTgHE9B",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df['airlines_name']=df['airlines_name'].apply(lambda x: 'A' if x=='6c483c0812c96f8ec43bb0ff76eaf716' else 'B' if x=='33199710eb822fbcfd0dc793f4788d30' else 'C' if x=='0a102015e48c1f68e121acc99fca9a05' else \n",
        "                          'D' if x=='ad5bef60d81ea077018f4d50b813153a' else 'E' if x=='74c5549aa99d55280a896ea50068a211' else 'F' if x=='e35de6a36d385711a660c72c0286154a' else 'G' if x=='9855a1d3de1c46526dde37c5d6fb758c' else 'H')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-nsgRBEyHE9G",
        "colab_type": "text"
      },
      "source": [
        "Based on previous analysis, G and H airlines will be removed from data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fKbvEvEhHE9H",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df=df[(df['airlines_name']!='G') & (df['airlines_name']!='H')]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OmOrRwjLHE9M",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**account_id** with too many order count will be removed from data\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "jupyter": {
          "outputs_hidden": true
        },
        "id": "WUQJ7IMjHE9N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df=df[(df['account_id']!='d64a90a618202a5e8b25d8539377f3ca')]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oz3bIZweHE9R",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "**route** constant values\n",
        "\n",
        "Because all of the observation in the data is the flight with route Jakarta to Denpasar, then **route** variable is not used in the modelling"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "M33WQEQp3KPx",
        "colab": {}
      },
      "source": [
        "df=df.reset_index()\n",
        "df=df.drop(['route','index'],1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "GbnaxWEpQbbe"
      },
      "source": [
        "# Feature Engineering \n",
        "\n",
        "The new Features created are as follows:\n",
        "* **is_refund** , an indicator of refunded transaction. We are assuming that negative transaction means a refunded transaction\n",
        "* Statistics from  **log_transaction** including mean, standard deviation, count,min,max\n",
        "* **visited_[city]**, an indicator if the user has visited that city or not. We are not creating **visited_medan**,**visited_denpasar**,**visited_jakarta** variable because all of the users has visited those cities\n",
        "* **adj_price** , is the result of  **price**  divided by  **no_of_seat**. If the **trip** is classified as *round* trip, then it will be divided by 2\n",
        "* **price_dif** , is the difference between **adj_price** and  **price** average given the same **service_class** and **airline_names**\n",
        "\n",
        "There are also some feaures which created when doing the cross validation to prevent overfitting:\n",
        "* **Woe** (Weight of Evidence) , is the intention of some group did cross selling. $$WoE = ln(\\frac{Prob(Yes)}{Prob(Np)})$$ . \n",
        "* **Adjusted_Woe**, is the modified version of **WoE** which also take account the number of observation\n",
        "* **mean_encoding** , is the probability of some group did cross selling\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "1KdabyafzJ99",
        "jupyter": {
          "outputs_hidden": true
        },
        "outputId": "da57874c-1248-4ed3-cfd2-156b0a564eca",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "df['log_transaction']=df['log_transaction'].apply(lambda x: ast.literal_eval(x))\n",
        "df['log_transaction']=df['log_transaction'].apply(lambda x: ast.literal_eval(x))\n",
        "print('Checking Negative Trx: ')\n",
        "print(df['log_transaction'].apply(lambda x: \"Negative\" if any(k<0 for k in x) else \"Positive\").value_counts())\n",
        "df['is_refund']=df['log_transaction'].apply(lambda x: \"Yes\" if any(k<0 for k in x) else \"No\")\n",
        "df['min_trx']=df['log_transaction'].apply(lambda x: np.min(x))\n",
        "df['max_trx']=df['log_transaction'].apply(lambda x: np.max(x))\n",
        "df['avg_trx']=df['log_transaction'].apply(lambda x: np.mean(x))\n",
        "df['std_trx']=df['log_transaction'].apply(lambda x: np.std(x))\n",
        "df['cnt_trx']=df['log_transaction'].apply(lambda x: len(x))\n",
        "# df['skew_trx']=df['log_transaction'].apply(lambda x: skew(x))\n",
        "\n",
        "df['visited_jogja']=df['visited_city'].apply(lambda x: 1 if 'Jogjakarta' in x else 0)\n",
        "df['visited_semarang']=df['visited_city'].apply(lambda x: 1 if 'Semarang' in x else 0)\n",
        "df['visited_surabaya']=df['visited_city'].apply(lambda x: 1 if 'Surabaya' in x else 0)\n",
        "df['visited_aceh']=df['visited_city'].apply(lambda x: 1 if 'Aceh' in x else 0)\n",
        "df['visited_manado']=df['visited_city'].apply(lambda x: 1 if 'Manado' in x else 0)\n",
        "\n",
        "df['adj_price']=df['price']/(df['no_of_seats'])\n",
        "df['adj_price']=df.apply(lambda x: (x.adj_price)/2 if x.trip=='round' else x.adj_price ,axis=1)\n",
        "\n",
        "an = df.airlines_name.unique()\n",
        "sc = df.service_class.unique()\n",
        "a=dict()\n",
        "for i in an:\n",
        "  for j in sc:\n",
        "    mean_g = np.mean(df[(df.airlines_name==i)&(df.service_class==j)&(df.is_tx_promo=='NO')]['adj_price'])\n",
        "    a[i+'_'+j]=mean_g\n",
        "df['price_dif']=df.apply(lambda x : ((x.adj_price) - a['A_ECONOMY']) if ((x.airlines_name=='A') & (x.service_class=='ECONOMY')) else \n",
        "                         ((x.adj_price) - a['B_ECONOMY']) if ((x.airlines_name=='B') & (x.service_class=='ECONOMY')) else \n",
        "                         ((x.adj_price) - a['C_ECONOMY']) if ((x.airlines_name=='C') & (x.service_class=='ECONOMY')) else \n",
        "                         ((x.adj_price) - a['D_ECONOMY']) if ((x.airlines_name=='D') & (x.service_class=='ECONOMY')) else \n",
        "                         ((x.adj_price) - a['E_BUSINESS']) if ((x.airlines_name=='E') & (x.service_class=='BUSINESS')) else \n",
        "                         ((x.adj_price) - a['E_ECONOMY']) if ((x.airlines_name=='E') & (x.service_class=='ECONOMY')) else \n",
        "                         ((x.adj_price) - a['F_ECONOMY']) if ((x.airlines_name=='F') & (x.service_class=='ECONOMY')) else \n",
        "                         np.nan ,axis=1)\n",
        "\n",
        "df=df.drop(['log_transaction','visited_city','hotel_id'],1)\n",
        "df_raw=df.copy()\n",
        "df=df.drop('order_id',1)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Checking Negative Trx: \n",
            "Positive    114540\n",
            "Negative        87\n",
            "Name: log_transaction, dtype: int64\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "F-8ccB960Zjl",
        "outputId": "82cc1583-ac22-424a-e34f-6881c0747f16",
        "colab": {}
      },
      "source": [
        "df.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>account_id</th>\n",
              "      <th>member_duration_days</th>\n",
              "      <th>gender</th>\n",
              "      <th>trip</th>\n",
              "      <th>service_class</th>\n",
              "      <th>price</th>\n",
              "      <th>is_tx_promo</th>\n",
              "      <th>no_of_seats</th>\n",
              "      <th>airlines_name</th>\n",
              "      <th>is_cross_sell</th>\n",
              "      <th>...</th>\n",
              "      <th>avg_trx</th>\n",
              "      <th>std_trx</th>\n",
              "      <th>cnt_trx</th>\n",
              "      <th>visited_jogja</th>\n",
              "      <th>visited_semarang</th>\n",
              "      <th>visited_surabaya</th>\n",
              "      <th>visited_aceh</th>\n",
              "      <th>visited_manado</th>\n",
              "      <th>adj_price</th>\n",
              "      <th>price_dif</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>912aa410a02cd7e1bab414214a7005c0</td>\n",
              "      <td>566.0</td>\n",
              "      <td>M</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>885898.0</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>B</td>\n",
              "      <td>No</td>\n",
              "      <td>...</td>\n",
              "      <td>1.030959e+06</td>\n",
              "      <td>4.533539e+05</td>\n",
              "      <td>4</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>885898.0</td>\n",
              "      <td>-3.502775e+05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1a42ac02bcb4a902973123323f84da55</td>\n",
              "      <td>648.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>2695550.0</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>C</td>\n",
              "      <td>No</td>\n",
              "      <td>...</td>\n",
              "      <td>4.489474e+06</td>\n",
              "      <td>2.250021e+06</td>\n",
              "      <td>5</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>2695550.0</td>\n",
              "      <td>1.382226e+06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>92cddd64d4be4dec6dfbcc0c50e902f4</td>\n",
              "      <td>418.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>1146665.0</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>C</td>\n",
              "      <td>No</td>\n",
              "      <td>...</td>\n",
              "      <td>2.954880e+06</td>\n",
              "      <td>1.706745e+06</td>\n",
              "      <td>3</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1146665.0</td>\n",
              "      <td>-1.666595e+05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>bf637abc47ea93bad22264f4956d67f6</td>\n",
              "      <td>537.0</td>\n",
              "      <td>F</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>1131032.5</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>A</td>\n",
              "      <td>No</td>\n",
              "      <td>...</td>\n",
              "      <td>4.362199e+06</td>\n",
              "      <td>3.006539e+06</td>\n",
              "      <td>157</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1131032.5</td>\n",
              "      <td>-1.017806e+05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>fdd499437b284d3c410ba3dfba4fe810</td>\n",
              "      <td>493.0</td>\n",
              "      <td>M</td>\n",
              "      <td>trip</td>\n",
              "      <td>ECONOMY</td>\n",
              "      <td>1457118.0</td>\n",
              "      <td>NO</td>\n",
              "      <td>1.0</td>\n",
              "      <td>D</td>\n",
              "      <td>No</td>\n",
              "      <td>...</td>\n",
              "      <td>5.302909e+06</td>\n",
              "      <td>2.723540e+06</td>\n",
              "      <td>3</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1457118.0</td>\n",
              "      <td>2.025159e+05</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 23 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                         account_id  member_duration_days gender  trip  \\\n",
              "0  912aa410a02cd7e1bab414214a7005c0                 566.0      M  trip   \n",
              "1  1a42ac02bcb4a902973123323f84da55                 648.0      F  trip   \n",
              "2  92cddd64d4be4dec6dfbcc0c50e902f4                 418.0      F  trip   \n",
              "3  bf637abc47ea93bad22264f4956d67f6                 537.0      F  trip   \n",
              "4  fdd499437b284d3c410ba3dfba4fe810                 493.0      M  trip   \n",
              "\n",
              "  service_class      price is_tx_promo  no_of_seats airlines_name  \\\n",
              "0       ECONOMY   885898.0          NO          1.0             B   \n",
              "1       ECONOMY  2695550.0          NO          1.0             C   \n",
              "2       ECONOMY  1146665.0          NO          1.0             C   \n",
              "3       ECONOMY  1131032.5          NO          1.0             A   \n",
              "4       ECONOMY  1457118.0          NO          1.0             D   \n",
              "\n",
              "  is_cross_sell  ...       avg_trx       std_trx  cnt_trx  visited_jogja  \\\n",
              "0            No  ...  1.030959e+06  4.533539e+05        4              0   \n",
              "1            No  ...  4.489474e+06  2.250021e+06        5              0   \n",
              "2            No  ...  2.954880e+06  1.706745e+06        3              1   \n",
              "3            No  ...  4.362199e+06  3.006539e+06      157              1   \n",
              "4            No  ...  5.302909e+06  2.723540e+06        3              0   \n",
              "\n",
              "   visited_semarang  visited_surabaya  visited_aceh  visited_manado  \\\n",
              "0                 1                 0             0               0   \n",
              "1                 1                 0             0               0   \n",
              "2                 0                 0             0               0   \n",
              "3                 1                 0             0               0   \n",
              "4                 0                 0             0               0   \n",
              "\n",
              "   adj_price     price_dif  \n",
              "0   885898.0 -3.502775e+05  \n",
              "1  2695550.0  1.382226e+06  \n",
              "2  1146665.0 -1.666595e+05  \n",
              "3  1131032.5 -1.017806e+05  \n",
              "4  1457118.0  2.025159e+05  \n",
              "\n",
              "[5 rows x 23 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 44
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "45qoDMCvj-oB"
      },
      "source": [
        "# Modelling\n",
        "\n",
        "Here is the modelling flow:\n",
        "1. Determining the combination of variables used in the modelling\n",
        "2. Determining which ML model want to be used\n",
        "3. Finding the best base parameter using the help of automl function provided by H2O library  \n",
        "4. Cross-validation manually equipped with added feature engineering\n",
        "\n",
        "After conduction the experiment of various variables combination, we end up of using this combination:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "5tfwlp9OkDkM",
        "outputId": "c5a649c0-cbd5-4bfe-e51e-c2a6b5c57d1f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        }
      },
      "source": [
        "drop_columns=['service_class','gender','price','is_refund','visited_jogja','visited_semarang','visited_surabaya','visited_aceh','visited_manado']\n",
        "df.drop(columns=drop_columns).columns"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['account_id', 'member_duration_days', 'trip', 'is_tx_promo',\n",
              "       'no_of_seats', 'airlines_name', 'is_cross_sell', 'min_trx', 'max_trx',\n",
              "       'avg_trx', 'std_trx', 'cnt_trx', 'adj_price', 'price_dif'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5AvQEiWtHE9f",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "## Model 1"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "baOnhhULHE9g",
        "colab_type": "text"
      },
      "source": [
        "Here we find the best base parameter for Random Forest model with the help of automl function provided by H2O library"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "tQMpxWYVkAHA",
        "outputId": "6ad26642-a6d6-461d-f14d-4d57b456cc0f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 542
        }
      },
      "source": [
        "#Initialize H2O\n",
        "h2o.init()\n",
        "\n",
        "drop_columns=['service_class','gender','price','is_refund','visited_jogja','visited_semarang','visited_surabaya','visited_aceh','visited_manado']\n",
        "drop_columns.append('account_id')\n",
        "\n",
        "#convert to H2OFrame\n",
        "df_h2o = h2o.H2OFrame(df.drop(drop_columns,1))\n",
        "\n",
        "#specify the target and predictor variable\n",
        "except_target=drop_columns.copy()\n",
        "except_target.append('is_cross_sell')\n",
        "x=df.drop(except_target,1).columns.tolist()\n",
        "y='is_cross_sell'\n",
        "\n",
        "#using automl with XGBoost Model to help find the best parameter\n",
        "aml_init = H2OAutoML(max_runtime_secs=3000, balance_classes=False,seed = 1,nfolds=5,stopping_rounds=5,\n",
        "               sort_metric='logloss',include_algos=['DRF'])\n",
        "aml_init.train(x = x, y = y, training_frame = df_h2o)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Checking whether there is an H2O instance running at http://localhost:54321 ..... not found.\n",
            "Attempting to start a local H2O server...\n",
            "  Java Version: openjdk version \"1.8.0_232\"; OpenJDK Runtime Environment (build 1.8.0_232-8u232-b09-1~deb9u1-b09); OpenJDK 64-Bit Server VM (build 25.232-b09, mixed mode)\n",
            "  Starting server from /opt/conda/lib/python3.6/site-packages/h2o/backend/bin/h2o.jar\n",
            "  Ice root: /tmp/tmp2gvukfgl\n",
            "  JVM stdout: /tmp/tmp2gvukfgl/h2o_unknownUser_started_from_python.out\n",
            "  JVM stderr: /tmp/tmp2gvukfgl/h2o_unknownUser_started_from_python.err\n",
            "  Server is running at http://127.0.0.1:54321\n",
            "Connecting to H2O server at http://127.0.0.1:54321 ... successful.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<div style=\"overflow:auto\"><table style=\"width:50%\"><tr><td>H2O cluster uptime:</td>\n",
              "<td>02 secs</td></tr>\n",
              "<tr><td>H2O cluster timezone:</td>\n",
              "<td>Etc/UTC</td></tr>\n",
              "<tr><td>H2O data parsing timezone:</td>\n",
              "<td>UTC</td></tr>\n",
              "<tr><td>H2O cluster version:</td>\n",
              "<td>3.26.0.10</td></tr>\n",
              "<tr><td>H2O cluster version age:</td>\n",
              "<td>1 month and 21 days </td></tr>\n",
              "<tr><td>H2O cluster name:</td>\n",
              "<td>H2O_from_python_unknownUser_8tzm01</td></tr>\n",
              "<tr><td>H2O cluster total nodes:</td>\n",
              "<td>1</td></tr>\n",
              "<tr><td>H2O cluster free memory:</td>\n",
              "<td>3.556 Gb</td></tr>\n",
              "<tr><td>H2O cluster total cores:</td>\n",
              "<td>4</td></tr>\n",
              "<tr><td>H2O cluster allowed cores:</td>\n",
              "<td>4</td></tr>\n",
              "<tr><td>H2O cluster status:</td>\n",
              "<td>accepting new members, healthy</td></tr>\n",
              "<tr><td>H2O connection url:</td>\n",
              "<td>http://127.0.0.1:54321</td></tr>\n",
              "<tr><td>H2O connection proxy:</td>\n",
              "<td>{'http': None, 'https': None}</td></tr>\n",
              "<tr><td>H2O internal security:</td>\n",
              "<td>False</td></tr>\n",
              "<tr><td>H2O API Extensions:</td>\n",
              "<td>Amazon S3, XGBoost, Algos, AutoML, Core V3, TargetEncoder, Core V4</td></tr>\n",
              "<tr><td>Python version:</td>\n",
              "<td>3.6.6 final</td></tr></table></div>"
            ],
            "text/plain": [
              "--------------------------  ------------------------------------------------------------------\n",
              "H2O cluster uptime:         02 secs\n",
              "H2O cluster timezone:       Etc/UTC\n",
              "H2O data parsing timezone:  UTC\n",
              "H2O cluster version:        3.26.0.10\n",
              "H2O cluster version age:    1 month and 21 days\n",
              "H2O cluster name:           H2O_from_python_unknownUser_8tzm01\n",
              "H2O cluster total nodes:    1\n",
              "H2O cluster free memory:    3.556 Gb\n",
              "H2O cluster total cores:    4\n",
              "H2O cluster allowed cores:  4\n",
              "H2O cluster status:         accepting new members, healthy\n",
              "H2O connection url:         http://127.0.0.1:54321\n",
              "H2O connection proxy:       {'http': None, 'https': None}\n",
              "H2O internal security:      False\n",
              "H2O API Extensions:         Amazon S3, XGBoost, Algos, AutoML, Core V3, TargetEncoder, Core V4\n",
              "Python version:             3.6.6 final\n",
              "--------------------------  ------------------------------------------------------------------"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "AutoML progress: |████████████████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "2Jr9JzAbmJ9n",
        "outputId": "96262b4c-2486-4363-b47e-1cd4c103fd71",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 99
        }
      },
      "source": [
        "#Leaderboard\n",
        "lb=aml_init.leaderboard\n",
        "print(lb)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table>\n",
              "<thead>\n",
              "<tr><th>model_id                    </th><th style=\"text-align: right;\">     auc</th><th style=\"text-align: right;\">  logloss</th><th style=\"text-align: right;\">  mean_per_class_error</th><th style=\"text-align: right;\">    rmse</th><th style=\"text-align: right;\">      mse</th></tr>\n",
              "</thead>\n",
              "<tbody>\n",
              "<tr><td>DRF_1_AutoML_20191228_195157</td><td style=\"text-align: right;\">0.709269</td><td style=\"text-align: right;\"> 0.205704</td><td style=\"text-align: right;\">              0.399334</td><td style=\"text-align: right;\">0.226093</td><td style=\"text-align: right;\">0.0511179</td></tr>\n",
              "<tr><td>XRT_1_AutoML_20191228_195157</td><td style=\"text-align: right;\">0.70957 </td><td style=\"text-align: right;\"> 0.20709 </td><td style=\"text-align: right;\">              0.406207</td><td style=\"text-align: right;\">0.226194</td><td style=\"text-align: right;\">0.0511637</td></tr>\n",
              "</tbody>\n",
              "</table>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "tBdMqpWumMO4",
        "outputId": "9b508e51-1006-47f1-8cac-9e2e0c828031",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        }
      },
      "source": [
        "#Variable Importance of Leader Model\n",
        "m = h2o.get_model(lb[0,\"model_id\"])\n",
        "df_imp=m.varimp(use_pandas=True)\n",
        "df_imp.index=df_imp['variable']\n",
        "\n",
        "df_imp['scaled_importance'].sort_values().plot.barh(figsize=(15,8))\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "d8Wcs6HBtivP",
        "colab": {}
      },
      "source": [
        "#DRF\n",
        "params_DRF = {\"ntrees\" : aml_init.leader.params['ntrees']['actual']\n",
        "    , \"max_depth\" : aml_init.leader.params['max_depth']['actual']\n",
        "    , \"keep_cross_validation_predictions\" : aml_init.leader.params['keep_cross_validation_predictions']['actual']\n",
        "    , \"nfolds\" : aml_init.leader.params['nfolds']['actual']\n",
        "    , \"nbins\" : aml_init.leader.params['nbins']['actual']\n",
        "    , \"nbins_cats\" : aml_init.leader.params['nbins_cats']['actual']    \n",
        "    , \"nbins_top_level\" : aml_init.leader.params['nbins_top_level']['actual']\n",
        "    # , \"learn_rate_annealing\" : aml_init.leader.params['learn_rate_annealing']['actual']\n",
        "    , \"distribution\" : aml_init.leader.params['distribution']['actual']\n",
        "    , \"sample_rate\" : aml_init.leader.params['sample_rate']['actual']\n",
        "    , \"sample_rate_per_class\" : aml_init.leader.params['sample_rate_per_class']['actual']\n",
        "    # , \"col_sample_rate\" :aml_init.leader.params['col_sample_rate']['actual']\n",
        "    , \"col_sample_rate_change_per_level\" : aml_init.leader.params['col_sample_rate_change_per_level']['actual']\n",
        "    , \"col_sample_rate_per_tree\" : aml_init.leader.params['col_sample_rate_per_tree']['actual']\n",
        "    , \"min_rows\" : aml_init.leader.params['min_rows']['actual']\n",
        "    # , \"max_abs_leafnode_pred\" : aml_init.leader.params['max_abs_leafnode_pred']['actual']\n",
        "    # , \"pred_noise_bandwidth\" : aml_init.leader.params['pred_noise_bandwidth']['actual']\n",
        "    , \"categorical_encoding\" : aml_init.leader.params['categorical_encoding']['actual']\n",
        "    # , \"min_child_weight\" :aml_init.leader.params['min_child_weight']['actual']\n",
        "    , \"seed\": aml_init.leader.params['seed']['actual']\n",
        "    , \"score_tree_interval\": aml_init.leader.params['score_tree_interval']['actual']\n",
        "    # , \"subsample\" :aml_init.leader.params['subsample']['actual']\n",
        "    # , \"max_bins\" :aml_init.leader.params['max_bins']['actual']\n",
        "    # , \"max_delta_step\" :aml_init.leader.params['max_delta_step']['actual']\n",
        "    # , \"max_leaves\" :aml_init.leader.params['max_leaves']['actual']\n",
        "    # , \"booster\" :aml_init.leader.params['booster']['actual']\n",
        "    # , \"sample_type\" :aml_init.leader.params['sample_type']['actual']\n",
        "    , \"stopping_rounds\" :aml_init.leader.params['stopping_rounds']['actual']\n",
        "    , \"stopping_metric\" :aml_init.leader.params['stopping_metric']['actual']\n",
        "    , \"stopping_tolerance\" :aml_init.leader.params['stopping_tolerance']['actual']\n",
        "    # , \"nthread\" :aml_init.leader.params['nthread']['actual']\n",
        "    # , \"eta\" :aml_init.leader.params['eta']['actual']\n",
        "    , \"min_split_improvement\" :aml_init.leader.params['min_split_improvement']['actual']\n",
        "    , \"histogram_type\" : aml_init.leader.params['histogram_type']['actual']\n",
        "    , \"score_each_iteration\" : aml_init.leader.params['score_each_iteration']['actual']\n",
        "    # , \"fold_assignment\" : aml_init.leader.params['fold_assignment']['actual']\n",
        "    # , \"fold_column\" : aml_init.leader.params['fold_column']['actual']\n",
        "    # , \"offset_column\" : aml_init.leader.params['offset_column']['actual']\n",
        "    , \"max_runtime_secs\" : aml_init.leader.params['max_runtime_secs']['actual']\n",
        "    , \"max_hit_ratio_k\" : aml_init.leader.params['max_hit_ratio_k']['actual']\n",
        "    , \"r2_stopping\" : aml_init.leader.params['r2_stopping']['actual']\n",
        "    , \"stopping_rounds\" : aml_init.leader.params['stopping_rounds']['actual']\n",
        "    , \"stopping_metric\" : aml_init.leader.params['stopping_metric']['actual']\n",
        "    , \"stopping_tolerance\" : aml_init.leader.params['max_runtime_secs']['actual']\n",
        "    , \"build_tree_one_node\" : aml_init.leader.params['build_tree_one_node']['actual']\n",
        "    , \"binomial_double_trees\" : aml_init.leader.params['binomial_double_trees']['actual']\n",
        "    # , \"tweedie_power\" : aml_init.leader.params['tweedie_power']['actual']\n",
        "    # , \"huber_alpha\" : aml_init.leader.params['huber_alpha']['actual']\n",
        "    , \"class_sampling_factors\" : aml_init.leader.params['class_sampling_factors']['actual']\n",
        "    , \"max_after_balance_size\" : aml_init.leader.params['max_after_balance_size']['actual']\n",
        "    , \"nbins_top_level\" : aml_init.leader.params['nbins_top_level']['actual']\n",
        "    , \"calibrate_model\" : aml_init.leader.params['calibrate_model']['actual']\n",
        "    , \"mtries\" : aml_init.leader.params['mtries']['actual']\n",
        "    , \"custom_metric_func\" : aml_init.leader.params['custom_metric_func']['actual']\n",
        "    # , \"custom_distribution_func\" : aml_init.leader.params['custom_distribution_func']['actual']\n",
        "    , \"export_checkpoints_dir\" : aml_init.leader.params['export_checkpoints_dir']['actual']\n",
        "    # , \"monotone_constraints\" : aml_init.leader.params['monotone_constraints']['actual']\n",
        "    , \"check_constant_response\" : aml_init.leader.params['check_constant_response']['actual']\n",
        "    # , \"reg_lambda\" :aml_init.leader.params['reg_lambda']['actual']\n",
        "    # , \"reg?_alpha\" :aml_init.leader.params['reg_alpha']['actual']\n",
        "}\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "03tjkdoJHE9w",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "## Model 2"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vSYp6yCJHE9x",
        "colab_type": "text"
      },
      "source": [
        "Here we find the best base parameter for GBM with the help of automl function provided by H2O library"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n6QTmL2UHE9y",
        "colab_type": "code",
        "outputId": "124487e3-09ff-4e15-d8c4-24f1096b0204",
        "colab": {}
      },
      "source": [
        "#Initialize H2O\n",
        "h2o.init()\n",
        "\n",
        "drop_columns=['service_class','gender','price','is_refund','visited_jogja','visited_semarang','visited_surabaya','visited_aceh','visited_manado']\n",
        "drop_columns.append('account_id')\n",
        "\n",
        "#convert to H2OFrame\n",
        "df_h2o = h2o.H2OFrame(df.drop(drop_columns,1))\n",
        "\n",
        "#specify the target and predictor variable\n",
        "except_target=drop_columns.copy()\n",
        "except_target.append('is_cross_sell')\n",
        "x=df.drop(except_target,1).columns.tolist()\n",
        "y='is_cross_sell'\n",
        "\n",
        "#using automl with XGBoost Model to help find the best parameter\n",
        "aml2_init = H2OAutoML(max_runtime_secs=3000, balance_classes=False,seed = 1,nfolds=5,stopping_rounds=5,\n",
        "               sort_metric='logloss',include_algos=['GBM'])\n",
        "aml2_init.train(x = x, y = y, training_frame = df_h2o)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Checking whether there is an H2O instance running at http://localhost:54321 . connected.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<div style=\"overflow:auto\"><table style=\"width:50%\"><tr><td>H2O cluster uptime:</td>\n",
              "<td>3 mins 36 secs</td></tr>\n",
              "<tr><td>H2O cluster timezone:</td>\n",
              "<td>Etc/UTC</td></tr>\n",
              "<tr><td>H2O data parsing timezone:</td>\n",
              "<td>UTC</td></tr>\n",
              "<tr><td>H2O cluster version:</td>\n",
              "<td>3.26.0.10</td></tr>\n",
              "<tr><td>H2O cluster version age:</td>\n",
              "<td>1 month and 21 days </td></tr>\n",
              "<tr><td>H2O cluster name:</td>\n",
              "<td>H2O_from_python_unknownUser_8tzm01</td></tr>\n",
              "<tr><td>H2O cluster total nodes:</td>\n",
              "<td>1</td></tr>\n",
              "<tr><td>H2O cluster free memory:</td>\n",
              "<td>3.320 Gb</td></tr>\n",
              "<tr><td>H2O cluster total cores:</td>\n",
              "<td>4</td></tr>\n",
              "<tr><td>H2O cluster allowed cores:</td>\n",
              "<td>4</td></tr>\n",
              "<tr><td>H2O cluster status:</td>\n",
              "<td>locked, healthy</td></tr>\n",
              "<tr><td>H2O connection url:</td>\n",
              "<td>http://localhost:54321</td></tr>\n",
              "<tr><td>H2O connection proxy:</td>\n",
              "<td>{'http': None, 'https': None}</td></tr>\n",
              "<tr><td>H2O internal security:</td>\n",
              "<td>False</td></tr>\n",
              "<tr><td>H2O API Extensions:</td>\n",
              "<td>Amazon S3, XGBoost, Algos, AutoML, Core V3, TargetEncoder, Core V4</td></tr>\n",
              "<tr><td>Python version:</td>\n",
              "<td>3.6.6 final</td></tr></table></div>"
            ],
            "text/plain": [
              "--------------------------  ------------------------------------------------------------------\n",
              "H2O cluster uptime:         3 mins 36 secs\n",
              "H2O cluster timezone:       Etc/UTC\n",
              "H2O data parsing timezone:  UTC\n",
              "H2O cluster version:        3.26.0.10\n",
              "H2O cluster version age:    1 month and 21 days\n",
              "H2O cluster name:           H2O_from_python_unknownUser_8tzm01\n",
              "H2O cluster total nodes:    1\n",
              "H2O cluster free memory:    3.320 Gb\n",
              "H2O cluster total cores:    4\n",
              "H2O cluster allowed cores:  4\n",
              "H2O cluster status:         locked, healthy\n",
              "H2O connection url:         http://localhost:54321\n",
              "H2O connection proxy:       {'http': None, 'https': None}\n",
              "H2O internal security:      False\n",
              "H2O API Extensions:         Amazon S3, XGBoost, Algos, AutoML, Core V3, TargetEncoder, Core V4\n",
              "Python version:             3.6.6 final\n",
              "--------------------------  ------------------------------------------------------------------"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "AutoML progress: |████████████████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nDdY_OkmHE93",
        "colab_type": "code",
        "outputId": "1ed12a33-a62d-4b9d-a881-e32234aa7dba",
        "colab": {}
      },
      "source": [
        "#Leaderboard\n",
        "lb=aml_init.leaderboard\n",
        "print(lb)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table>\n",
              "<thead>\n",
              "<tr><th>model_id                    </th><th style=\"text-align: right;\">     auc</th><th style=\"text-align: right;\">  logloss</th><th style=\"text-align: right;\">  mean_per_class_error</th><th style=\"text-align: right;\">    rmse</th><th style=\"text-align: right;\">      mse</th></tr>\n",
              "</thead>\n",
              "<tbody>\n",
              "<tr><td>DRF_1_AutoML_20191228_195157</td><td style=\"text-align: right;\">0.709269</td><td style=\"text-align: right;\"> 0.205704</td><td style=\"text-align: right;\">              0.399334</td><td style=\"text-align: right;\">0.226093</td><td style=\"text-align: right;\">0.0511179</td></tr>\n",
              "<tr><td>XRT_1_AutoML_20191228_195157</td><td style=\"text-align: right;\">0.70957 </td><td style=\"text-align: right;\"> 0.20709 </td><td style=\"text-align: right;\">              0.406207</td><td style=\"text-align: right;\">0.226194</td><td style=\"text-align: right;\">0.0511637</td></tr>\n",
              "</tbody>\n",
              "</table>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OtB3P-LSHE-A",
        "colab_type": "code",
        "outputId": "dea687bb-3dbd-4aa0-82c0-d9eeedf3744f",
        "colab": {}
      },
      "source": [
        "#Variable Importance of Leader Model\n",
        "m = h2o.get_model(lb[0,\"model_id\"])\n",
        "df_imp=m.varimp(use_pandas=True)\n",
        "df_imp.index=df_imp['variable']\n",
        "\n",
        "df_imp['scaled_importance'].sort_values().plot.barh(figsize=(15,8))\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zPDZNTqPHE-D",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#gbm\n",
        "params_GBM = {\"ntrees\" : aml2_init.leader.params['ntrees']['actual']\n",
        "    , \"max_depth\" : aml2_init.leader.params['max_depth']['actual']\n",
        "    , \"keep_cross_validation_predictions\" : aml2_init.leader.params['keep_cross_validation_predictions']['actual']\n",
        "    , \"nfolds\" : aml2_init.leader.params['nfolds']['actual']\n",
        "    , \"nbins\" : aml2_init.leader.params['nbins']['actual']\n",
        "    , \"nbins_cats\" : aml2_init.leader.params['nbins_cats']['actual']\n",
        "    , \"learn_rate\" : aml2_init.leader.params['learn_rate']['actual']\n",
        "    , \"learn_rate_annealing\" : aml2_init.leader.params['learn_rate_annealing']['actual']\n",
        "    , \"distribution\" : aml2_init.leader.params['distribution']['actual']\n",
        "    , \"sample_rate\" : aml2_init.leader.params['sample_rate']['actual']\n",
        "    , \"sample_rate_per_class\" : aml2_init.leader.params['sample_rate_per_class']['actual']\n",
        "    , \"col_sample_rate\" :aml2_init.leader.params['col_sample_rate']['actual']\n",
        "    , \"col_sample_rate_change_per_level\" : aml2_init.leader.params['col_sample_rate_change_per_level']['actual']\n",
        "    , \"col_sample_rate_per_tree\" : aml2_init.leader.params['col_sample_rate_per_tree']['actual']\n",
        "    , \"min_rows\" : aml2_init.leader.params['min_rows']['actual']\n",
        "    , \"max_abs_leafnode_pred\" : aml2_init.leader.params['max_abs_leafnode_pred']['actual']\n",
        "    , \"pred_noise_bandwidth\" : aml2_init.leader.params['pred_noise_bandwidth']['actual']\n",
        "    , \"categorical_encoding\" : aml2_init.leader.params['categorical_encoding']['actual']\n",
        "    , \"seed\": aml2_init.leader.params['seed']['actual']\n",
        "    , \"score_tree_interval\": aml2_init.leader.params['score_tree_interval']['actual']\n",
        "    , \"stopping_rounds\" :aml2_init.leader.params['stopping_rounds']['actual']\n",
        "    , \"stopping_metric\" :aml2_init.leader.params['stopping_metric']['actual']\n",
        "    , \"stopping_tolerance\" :aml2_init.leader.params['stopping_tolerance']['actual']\n",
        "    , \"min_split_improvement\" :aml2_init.leader.params['min_split_improvement']['actual']\n",
        "    , \"histogram_type\" : aml2_init.leader.params['histogram_type']['actual']\n",
        "    , \"score_each_iteration\" : aml2_init.leader.params['score_each_iteration']['actual']\n",
        "    , \"max_hit_ratio_k\" : aml2_init.leader.params['max_hit_ratio_k']['actual']\n",
        "    , \"r2_stopping\" : aml2_init.leader.params['r2_stopping']['actual']\n",
        "    , \"stopping_rounds\" : aml2_init.leader.params['stopping_rounds']['actual']\n",
        "    , \"stopping_metric\" : aml2_init.leader.params['stopping_metric']['actual']\n",
        "    , \"stopping_tolerance\" : aml2_init.leader.params['max_runtime_secs']['actual']\n",
        "    , \"build_tree_one_node\" : aml2_init.leader.params['build_tree_one_node']['actual']\n",
        "    , \"quantile_alpha\" : aml2_init.leader.params['quantile_alpha']['actual']\n",
        "    , \"tweedie_power\" : aml2_init.leader.params['tweedie_power']['actual']\n",
        "    , \"huber_alpha\" : aml2_init.leader.params['huber_alpha']['actual']\n",
        "    , \"class_sampling_factors\" : aml2_init.leader.params['class_sampling_factors']['actual']\n",
        "    , \"max_after_balance_size\" : aml2_init.leader.params['max_after_balance_size']['actual']\n",
        "    , \"nbins_top_level\" : aml2_init.leader.params['nbins_top_level']['actual']\n",
        "    , \"calibrate_model\" : aml2_init.leader.params['calibrate_model']['actual']\n",
        "    , \"custom_metric_func\" : aml2_init.leader.params['custom_metric_func']['actual']\n",
        "    , \"custom_distribution_func\" : aml2_init.leader.params['custom_distribution_func']['actual']\n",
        "    , \"export_checkpoints_dir\" : aml2_init.leader.params['export_checkpoints_dir']['actual']\n",
        "    , \"monotone_constraints\" : aml2_init.leader.params['monotone_constraints']['actual']\n",
        "    , \"check_constant_response\" : aml2_init.leader.params['check_constant_response']['actual']\n",
        "}\n",
        "\n",
        "params_GBM['seed'] = params_DRF['seed']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PTPq28-aHE-I",
        "colab_type": "text"
      },
      "source": [
        "---\n",
        "## K-Fold Cross Validation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0Xi3i9auHE-J",
        "colab_type": "text"
      },
      "source": [
        "Below is another function definition for K-Fold cross validation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "w52DykTCMqC9",
        "colab": {}
      },
      "source": [
        "def recall_m(y_true, y_pred):\n",
        "  '''\n",
        "  Function to compute the recall metrics for keras model\n",
        "  '''\n",
        "  true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "  possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))\n",
        "  recall = true_positives / (possible_positives + K.epsilon())\n",
        "  return recall\n",
        "\n",
        "\n",
        "def precision_m(y_true, y_pred):\n",
        "  '''\n",
        "  Function to compute the precision metrics for keras model\n",
        "  '''\n",
        "  true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))\n",
        "  predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))\n",
        "  precision = true_positives / (predicted_positives + K.epsilon())\n",
        "  return precision\n",
        "\n",
        "\n",
        "def f1_m(y_true, y_pred):\n",
        "  '''\n",
        "  Function to compute the f1 metrics for keras model\n",
        "  '''\n",
        "  precision = precision_m(y_true, y_pred)\n",
        "  recall = recall_m(y_true, y_pred)\n",
        "  return 2*((precision*recall)/(precision+recall+K.epsilon()))\n",
        "\n",
        "\n",
        "def embedding_transform(df_train,df_val,column,target,epochs):\n",
        "  '''\n",
        "  Function to create the embedding form of categorical features based on df_train\n",
        "  '''\n",
        "  train=df_train.copy()\n",
        "  val=df_val.copy()\n",
        "  train[target]=train[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  val[target]=val[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  #embedding categorical features\n",
        "  no_of_unique_cat  = train[column].nunique()\n",
        "  embedding_size =300\n",
        "  vocab  = no_of_unique_cat+1\n",
        "\n",
        "  input_cat = Input(shape=(1,))\n",
        "  embedding=Embedding(vocab, embedding_size, input_length=1)(input_cat)\n",
        "  embedding=Reshape(target_shape=(embedding_size,))(embedding)\n",
        "  x=Dense(5,activation='relu')(embedding)\n",
        "  x=Dense(2,activation='relu')(embedding)\n",
        "  output=Dense(1,activation='sigmoid')(x)\n",
        "  model = Model(input_cat, output)\n",
        "\n",
        "  model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[f1_m])\n",
        "  max_length = 1\n",
        "  encoded_docs = [one_hot(d, embedding_size) for d in train[column]]\n",
        "  padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post')\n",
        "\n",
        "  encoded_docs_val = [one_hot(d, embedding_size) for d in val[column]]\n",
        "  padded_docs_val = pad_sequences(encoded_docs_val, maxlen=max_length, padding='post')\n",
        "\n",
        "  model.fit(padded_docs,train[target].values, epochs=epochs, verbose=1,batch_size=2048\n",
        "            ,validation_data=(padded_docs_val,val[target].values))\n",
        "  df_train[column+'_embedding']=model.predict(padded_docs) \n",
        "  df_val[column+'_embedding']=model.predict(padded_docs_val)\n",
        "  df_train=df_train.drop(column,1)\n",
        "  df_val=df_val.drop(column,1)\n",
        "  return(df_train,df_val)\n",
        "\n",
        "\n",
        "def ordering_label(df,target,feat):\n",
        "  '''\n",
        "  Function for ordering categorical features based on Target feature\n",
        "  '''\n",
        "  data=df.copy()\n",
        "  data[target]=data[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  ordered_labels = data.groupby(feat).mean()[target].sort_values().index\n",
        "  ordinal_label = {k:i for i, k in enumerate(ordered_labels, 0)} \n",
        "  df[feat]=df[feat].map(ordinal_label)\n",
        "  return(df)\n",
        "\n",
        "\n",
        "def rescaling(df,columns,scale_type='Standard'):\n",
        "  '''\n",
        "  Function for Feature Scaling\n",
        "  '''\n",
        "  scale_type=scale_type.lower()\n",
        "  scaled_X=df.drop(columns,1)\n",
        "  X=df[columns]\n",
        "  \n",
        "  if scale_type=='minmax':\n",
        "    scaler=MinMaxScaler(feature_range=(0,1))\n",
        "  elif scale_type=='standard':\n",
        "    scaler=StandardScaler()\n",
        "\n",
        "  scaled_column=scaler.fit_transform(X)\n",
        "  scaled_column=pd.DataFrame(scaled_column,columns=columns)\n",
        "  for column in columns:\n",
        "    scaled_X[column]=scaled_column[column].tolist()\n",
        "  \n",
        "  return(scaled_X)\n",
        "\n",
        "\n",
        "def detailed_segmentation(df_train,df_val,columns,target,num_segments):\n",
        "  '''\n",
        "  Function for detailed segmentation\n",
        "    \n",
        "  Number of segments which is not the multiple of 2 will be round down to the nearest\n",
        "  whole number\n",
        "  '''\n",
        "  depth=math.floor(math.log2(num_segments))\n",
        "  X_train=df_train.drop(target,1)\n",
        "  y_train=df_train[[target]]\n",
        "  X_val=df_val.drop(target,1)\n",
        "\n",
        "  tree_model = DecisionTreeClassifier(max_depth=depth)\n",
        "    \n",
        "  for column in columns:\n",
        "    tree_model.fit(X_train[[column]], y_train)\n",
        "    df_train[column+'detailed_segment'] = tree_model.predict_proba(X_train[[column]])[:,1]\n",
        "    df_train=ordering_label(df_train,target,column+'detailed_segment')\n",
        "    df_val[column+'detailed_segment'] = tree_model.predict_proba(X_val[[column]])[:,1]\n",
        "    df_val=ordering_label(df_val,target,column+'detailed_segment')\n",
        "    df_train=df_train.drop(column,1)\n",
        "    df_val=df_val.drop(column,1)\n",
        "  return(df_train,df_val)\n",
        "\n",
        "\n",
        "def mean_encoding(df_train,df_val,columns,target,adjusted=False):\n",
        "  '''\n",
        "  Function to create mean encoding\n",
        "  '''\n",
        "  train=df_train.copy()\n",
        "  train[target]=train[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  for column in columns:\n",
        "    prob_df = pd.DataFrame(train.groupby([column])[target].mean())\n",
        "    if adjusted:\n",
        "      prob_mean = prob_df[target].to_dict()\n",
        "      train['mean']=train[column].map(prob_mean)\n",
        "      train.set_index(['order_id'],inplace=True)\n",
        "      ordered_labels_train=(train['mean']*train['cnt_trx']).to_dict()\n",
        "      df_train[column+'_adj_mean_encoding']= df_train['order_id'].map(ordered_labels_train)\n",
        "\n",
        "      val=df_val.copy()\n",
        "      val[target]=val[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "      #if there is no historical data before then probability of success is assumed same as in the training set\n",
        "      train_mean=df_train[target].value_counts()['Yes']/len(df_train[target])\n",
        "      val['mean']=val[column].apply(lambda x: prob_mean[x] if x in prob_mean else train_mean)\n",
        "      val.set_index(['order_id'],inplace=True)\n",
        "      ordered_labels_val=(val['mean']*val['cnt_trx']).to_dict()\n",
        "      df_val[column+'_adj_mean_encoding'] = df_val['order_id'].map(ordered_labels_val)\n",
        "      df_train=df_train.drop(column,1)\n",
        "      df_val=df_val.drop(column,1)\n",
        "    else:\n",
        "      prob_df['mean']=prob_df['is_cross_sell']\n",
        "      ordered_labels = prob_df['mean'].to_dict()\n",
        "      df_train[column+'_mean_encoding'] = df_train[column].map(ordered_labels)\n",
        "      #if there is no historical data before then probability of success is assumed same as in the training set\n",
        "      train_mean=df_train[target].value_counts()['Yes']/len(df_train[target])\n",
        "      df_val[column+'_mean_encoding'] = df_val[column].apply(lambda x: ordered_labels[x] if x in ordered_labels \n",
        "                                                            else train_mean)\n",
        "      df_train=df_train.drop(column,1)\n",
        "      df_val=df_val.drop(column,1)\n",
        "  return(df_train,df_val)\n",
        "\n",
        "\n",
        "def PMI(df_train,df_val,columns):\n",
        "  '''\n",
        "  Function to create Pointwise Mutual Information\n",
        "  '''\n",
        "  for column in columns:\n",
        "    toy=df_train.copy()\n",
        "    toy=toy.sort_values(by=['account_id'])\n",
        "    toy['is_cross_sell']=toy['is_cross_sell'].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "    MBA_df=pd.DataFrame(toy.groupby([column])['is_cross_sell'].mean())\n",
        "    toy_all=toy[['account_id']]\n",
        "    len_toy_all=len(toy_all)\n",
        "    toy_pos=toy[toy.is_cross_sell==1][['account_id']]\n",
        "    len_toy_pos=len(toy_pos)\n",
        "    acc_id_unique=pd.Series(MBA_df.index.tolist())\n",
        "    acc_id_pos_unique=pd.Series(MBA_df[MBA_df.is_cross_sell==1].index.tolist())\n",
        "    #Calculating probability of each account_id\n",
        "    a=toy_all['account_id'].value_counts()/len_toy_all\n",
        "    prob_acc_id=acc_id_unique.apply(lambda x: a[x])\n",
        "    #Calculating Confidence\n",
        "    b=toy_pos['account_id'].value_counts()/len_toy_pos\n",
        "    confidence=acc_id_unique.apply(lambda x: b[x] if x in acc_id_pos_unique else 0)\n",
        "    print('Done confidence')\n",
        "    #Assigning Values to Dataframe\n",
        "    MBA_df['lift']=confidence/prob_acc_id\n",
        "    # MBA_df['adj_PMI']=np.log(1+MBA_df['lift'])\n",
        "    MBA_df['adj_PMI']=MBA_df['lift']\n",
        "    adj_PMI=MBA_df['adj_PMI'].to_dict()\n",
        "\n",
        "    df_train[column+'_adj_PMI'] = df_train[column].map(adj_PMI)\n",
        "    #if there is no historical data before then probability of that particular account_id is 0 which implies lift=0, so that adj_PMI=0\n",
        "    df_val[column+'_adj_PMI'] = df_val[column].apply(lambda x: adj_PMI[x] if x in adj_PMI \n",
        "                                                           else 0)\n",
        "    df_train=df_train.drop(column,1)\n",
        "    df_val=df_val.drop(column,1)\n",
        "  return(df_train,df_val)    \n",
        "\n",
        "\n",
        "def WoE(df_train,df_val,columns,target,adjusted=False):\n",
        "  '''\n",
        "  Function to create Weight of Evidence\n",
        "  '''\n",
        "  train=df_train.copy()\n",
        "  train[target]=train[target].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  for column in columns:\n",
        "    a = pd.DataFrame(train.groupby([column])[target].mean())\n",
        "    if adjusted:\n",
        "      b=pd.DataFrame(train.groupby([column])['is_cross_sell'].count())\n",
        "      prob_df=a.merge(b,on=column)\n",
        "      prob_df=prob_df.rename(columns={\"is_cross_sell_x\": \"is_cross_sell\", \"is_cross_sell_y\": \"count\"})\n",
        "    else:\n",
        "      prob_df=a.copy()\n",
        "    prob_df['not_'+target] = 1-prob_df[target]\n",
        "    prob_df[prob_df[target] == 0] = 0.000001\n",
        "    # prob_df[prob_df['not_'+target] == 0] = 0.000001\n",
        "    if adjusted:\n",
        "      prob_df['WoE'] = np.log(1+prob_df['count'])*np.log(prob_df[target]/prob_df['not_'+target])\n",
        "    else:\n",
        "      prob_df['WoE'] = np.log(prob_df[target]/prob_df['not_'+target])\n",
        "    ordered_labels = prob_df['WoE'].to_dict()\n",
        "\n",
        "    df_train[column+'_WoE'] = df_train[column].map(ordered_labels)\n",
        "    #if there is no historical data before then probability of success is assumed same as in the training set\n",
        "    df_val[column+'_WoE'] = df_val[column].apply(lambda x: ordered_labels[x] if x in ordered_labels else \n",
        "                                                 np.log((0.000001+df_train[target].value_counts()['Yes'])/(df_train[target].value_counts()['No'])) if df_train[target].value_counts()['Yes']==0 else\n",
        "                                                  np.log((df_train[target].value_counts()['Yes'])/(0.000001+df_train[target].value_counts()['No'])))\n",
        "    df_train=df_train.drop(column,1)\n",
        "    df_val=df_val.drop(column,1)\n",
        "  return(df_train,df_val)\n",
        "\n",
        "\n",
        "def kmeans(df,columns,k,scaled=True,plot=True):\n",
        "  '''\n",
        "  Function for kMeans segmentation\n",
        "  '''\n",
        "  if scaled==False:\n",
        "    X=rescaling(df,columns,scale_type='Standard')\n",
        "  else:\n",
        "    X=df[columns].copy()\n",
        "    \n",
        "  kmeans=MiniBatchKMeans(n_clusters=k ,init='k-means++',random_state=100,verbose=False)\n",
        "  y=kmeans.fit(X)\n",
        "  df['label']=y.labels_.tolist()\n",
        "  \n",
        "  if plot==True:\n",
        "    if len(columns)>2:\n",
        "      pca = PCA(n_components=3)\n",
        "      df_pca = pca.fit_transform(X)\n",
        "      df_pca=pd.DataFrame(df_pca,columns=['pca_1','pca_2','pca_3'])\n",
        "      fig=plt.figure()\n",
        "      ax=plt.axes(projection=\"3d\")\n",
        "      scatter=ax.scatter3D(df_pca.loc[:,'pca_1'],df_pca.loc[:,'pca_2'],df_pca.loc[:,'pca_3'],c=y.labels_,cmap='hsv')\n",
        "      plt.colorbar(scatter)\n",
        "      plt.title(str(k)+'-Means Clustering')\n",
        "      plt.show()\n",
        "    else:\n",
        "      scatter=plt.scatter(df.loc[:,columns[0]],df.loc[:,columns[1]],c=y.labels_)\n",
        "      plt.xlabel(columns[0])\n",
        "      plt.ylabel(columns[1])\n",
        "      plt.title(str(k)+'-Means Clustering')\n",
        "      plt.show()\n",
        "  return(df)\n",
        "\n",
        "\n",
        "def segmentation(df_train,df_val,k=9):\n",
        "  '''\n",
        "  Function to do kmeans clustering and assign the label into both df_train and df_val\n",
        "  '''\n",
        "  \n",
        "  toy=df_train.copy()\n",
        "  toy['is_cross_sell']=toy['is_cross_sell'].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "  df_seg=pd.DataFrame(toy.groupby('account_id').mean()['is_cross_sell'])\n",
        "  df_seg['count']=toy.groupby('account_id').count()['member_duration_days']\n",
        "  df_seg=df_seg.rename({'member_duration_days':'count'})\n",
        "  df_seg=kmeans(df_seg[df_seg['count']<20],columns=['is_cross_sell','count'],k=k,scaled=False,plot=False)\n",
        "  df_seg=ordering_label(df_seg,target='is_cross_sell',feat='label')\n",
        "  acc_0=df_seg[df_seg.label==0].index.tolist()\n",
        "  acc_1=df_seg[df_seg.label==1].index.tolist()\n",
        "  acc_2=df_seg[df_seg.label==2].index.tolist()\n",
        "  acc_3=df_seg[df_seg.label==3].index.tolist()\n",
        "  acc_4=df_seg[df_seg.label==4].index.tolist()\n",
        "  acc_5=df_seg[df_seg.label==5].index.tolist()\n",
        "  acc_6=df_seg[df_seg.label==6].index.tolist()\n",
        "  acc_7=df_seg[df_seg.label==7].index.tolist()\n",
        "  acc_8=df_seg[df_seg.label==8].index.tolist()\n",
        "  acc_9=df_seg[df_seg['count']>=20].index.tolist()\n",
        "  df_train['segment']=df_train.account_id.apply(lambda x: 0 if x in acc_0 else 1 if x in acc_1 else \n",
        "                                                2 if x in acc_2 else 3 if x in acc_3 else 4 if x in acc_4 else 5 if x in acc_5 else \n",
        "                                                6 if x in acc_6 else 7 if x in acc_7 else 8 if x in acc_8 else 9)\n",
        "  df_val['segment']=df_val.account_id.apply(lambda x: 0 if x in acc_0 else 1 if x in acc_1 else \n",
        "                                                2 if x in acc_2 else 3 if x in acc_3 else 4 if x in acc_4 else 5 if x in acc_5 else \n",
        "                                                6 if x in acc_6 else 7 if x in acc_7 else 8 if x in acc_8 else 9 if x in acc_9 else\n",
        "                                                np.nan)\n",
        "  return(df_train,df_val)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "FcWvWonHv-PJ",
        "colab": {}
      },
      "source": [
        "def cross_val(df,drop_columns,f1_threshold,params,nfolds,model='drf',is_feat_eng=False,label_not_cross_sell=False,is_binning=True,is_WoE=False,adjusted_WoE=True,is_mean_encode=True,adjusted_mean=False,is_segment=False,is_PMI=False,is_embedding=False):\n",
        "  '''\n",
        "  Function to do cross validation for single model\n",
        "  '''\n",
        "\n",
        "  f1_list=[]\n",
        "  #Initiate StratifiedKFold Class\n",
        "  kfold = KFold(n_splits=nfolds, shuffle=True, random_state=500)\n",
        "  #Drop Columns\n",
        "  if len(drop_columns)>0:\n",
        "    df=df.drop(drop_columns,1)\n",
        "  #begin looping\n",
        "  for train,val in kfold.split(df):\n",
        "    df_train=df.loc[train,:]\n",
        "    df_val=df.loc[val,:]\n",
        "\n",
        "    #Feature Engineering\n",
        "    if is_feat_eng:\n",
        "      if label_not_cross_sell:\n",
        "        #Creating list of customer which has number of orders more than 50\n",
        "        acc=pd.DataFrame(df_train['account_id'].value_counts())\n",
        "        top_acc_list=acc[acc.account_id>50].index.tolist()\n",
        "        #Creating list of customer which never done cross selling\n",
        "        toy=df_train.copy()\n",
        "        toy['is_cross_sell']=toy['is_cross_sell'].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "        cross_sell_prop=pd.DataFrame(toy.groupby('account_id').mean()['is_cross_sell'])\n",
        "        no_cross_sell_list=cross_sell_prop[cross_sell_prop['is_cross_sell']==0].index.tolist()\n",
        "        #Creating  list of customer which has number of orders more than 50 and never done cross selling\n",
        "        not_cross_sell_account_id=df_train[(df_train['account_id'].isin(top_acc_list)) & (df_train['account_id'].isin(no_cross_sell_list))]['account_id'].unique().tolist()\n",
        "\n",
        "        df_train['label_not_cross_sell']=df_train['account_id'].apply(lambda x: 1 if x in not_cross_sell_account_id else 0)\n",
        "        df_val['label_not_cross_sell']=df_val['account_id'].apply(lambda x: 1 if x in not_cross_sell_account_id else 0)\n",
        "      if is_segment:\n",
        "        df_train,df_val=segmentation(df_train,df_val,k=4)\n",
        "      if is_PMI:\n",
        "        df_train,df_val=PMI(df_train,df_val,columns=['account_id'])\n",
        "      if is_WoE:\n",
        "        df_train,df_val=WoE(df_train,df_val,columns=['airlines_name'],target='is_cross_sell',adjusted=adjusted_WoE)\n",
        "      if is_mean_encode:\n",
        "        df_train,df_val=mean_encoding(df_train,df_val,columns=['account_id'],target='is_cross_sell',adjusted=adjusted_mean)\n",
        "      if is_binning:\n",
        "        df_train,df_val=detailed_segmentation(df_train,df_val,columns=['member_duration_days','price'],target='is_cross_sell',num_segments=5)\n",
        "      if is_embedding:\n",
        "         df_train,df_val=embedding_transform(df_train,df_val,column='order_id',target='is_cross_sell',epochs=50)\n",
        "      if 'account_id' in df_train.columns:\n",
        "        df_train,df_val=df_train.drop('account_id',1),df_val.drop('account_id',1)\n",
        "      if 'order_id' in df_train.columns:\n",
        "        df_train,df_val=df_train.drop('order_id',1),df_val.drop('order_id',1)\n",
        "      print('Done Feature Engineering!')\n",
        "    else:\n",
        "      df_train,df_val=df_train.drop(['account_id','order_id'],1),df_val.drop(['account_id','order_id'],1)\n",
        "\n",
        "    df_train_h2o = h2o.H2OFrame(df_train)\n",
        "    df_val_h2o = h2o.H2OFrame(df_val)\n",
        "    #specify the target and predictor variable\n",
        "    x=df_train.drop('is_cross_sell',1).columns.tolist()\n",
        "    y='is_cross_sell'\n",
        "\n",
        "    if model=='gbm':\n",
        "      #Initiate Model\n",
        "      model_gbm = H2OGradientBoostingEstimator(**params_GBM)\n",
        "      model_gbm.train(x = x, y = y, training_frame = df_train_h2o,validation_frame = df_val_h2o)\n",
        "      #Variable Importance Plot\n",
        "      df_imp=model_gbm.varimp(use_pandas=True)\n",
        "      df_imp.index=df_imp['variable']\n",
        "      df_imp['scaled_importance'].sort_values().plot.barh(figsize=(15,8))\n",
        "      plt.show()\n",
        "      #Data Validation Prediction\n",
        "      prediction_prob = model_gbm.predict(df_val_h2o)[:,2]\n",
        "      prediction_prob=prediction_prob.as_data_frame()['Yes']\n",
        "    elif model=='drf':\n",
        "      #Initiate Model\n",
        "      model_drf = H2ORandomForestEstimator(**params_DRF)\n",
        "      model_drf.train(x = x, y = y, training_frame = df_train_h2o,validation_frame = df_val_h2o)\n",
        "      #Variable Importance Plot\n",
        "      df_imp=model_drf.varimp(use_pandas=True)\n",
        "      df_imp.index=df_imp['variable']\n",
        "      df_imp['scaled_importance'].sort_values().plot.barh(figsize=(15,8))\n",
        "      plt.show()\n",
        "      prediction_prob = model_drf.predict(df_val_h2o)[:,2]\n",
        "      prediction_prob=prediction_prob.as_data_frame()['Yes']\n",
        "    #Test various number of threshold for classifying \"Yes\" or \"No\"\n",
        "    for k in f1_threshold :\n",
        "        prediction=prediction_prob.apply(lambda x: \"Yes\" if x>k else \"No\")\n",
        "        # calculate the new F1\n",
        "        precision=confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]/(confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]+confusion_matrix(df_val['is_cross_sell'],prediction)[0][1])\n",
        "        recall=confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]/(confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]+confusion_matrix(df_val['is_cross_sell'],prediction)[1][0])\n",
        "        f1=2*precision*recall/(precision+recall)\n",
        "#         print('Threshold : {}, F1 Score: {}'.format(k,f1))\n",
        "        f1_list.append(f1)\n",
        "  hasil = pd.DataFrame(np.array(f1_list).reshape(5,len(f1_threshold)).transpose())\n",
        "  hasil.index = f1_threshold\n",
        "  hasil['mean'] = hasil.mean(axis=1)\n",
        "  hasil['std']=hasil.std(axis=1)\n",
        "  return hasil\n",
        "\n",
        "def cross_val_ensemble(df,drop_columns,f1_threshold,nfolds,model1='DRF',model2='GBM',is_feat_eng=False,label_not_cross_sell=False,is_binning=True,is_WoE=False,adjusted_WoE=False,is_mean_encode=True,adjusted_mean=False,is_segment=False):\n",
        "  '''\n",
        "  Function to do cross validation for ensemble model\n",
        "  '''\n",
        "\n",
        "  f1_list=[]\n",
        "  #Initiate StratifiedKFold Class\n",
        "  kfold = KFold(n_splits=nfolds, shuffle=True, random_state=500)\n",
        "  #Drop Columns\n",
        "  if len(drop_columns)>0:\n",
        "    df=df.drop(drop_columns,1)\n",
        "  #begin looping\n",
        "  for train,val in kfold.split(df):\n",
        "    df_train=df.loc[train,:]\n",
        "    df_val=df.loc[val,:]\n",
        "\n",
        "    #Feature Engineering\n",
        "    if is_feat_eng:\n",
        "      if label_not_cross_sell:\n",
        "        #Creating list of customer which has number of orders more than 50\n",
        "        acc=pd.DataFrame(df_train['account_id'].value_counts())\n",
        "        top_acc_list=acc[acc.account_id>50].index.tolist()\n",
        "        #Creating list of customer which never done cross selling\n",
        "        toy=df_train.copy()\n",
        "        toy['is_cross_sell']=toy['is_cross_sell'].apply(lambda x: 1 if x=='Yes' else 0)\n",
        "        cross_sell_prop=pd.DataFrame(toy.groupby('account_id').mean()['is_cross_sell'])\n",
        "        no_cross_sell_list=cross_sell_prop[cross_sell_prop['is_cross_sell']==0].index.tolist()\n",
        "        #Creating  list of customer which has number of orders more than 50 and never done cross selling\n",
        "        not_cross_sell_account_id=df_train[(df_train['account_id'].isin(top_acc_list)) & (df_train['account_id'].isin(no_cross_sell_list))]['account_id'].unique().tolist()\n",
        "\n",
        "        df_train['label_not_cross_sell']=df_train['account_id'].apply(lambda x: 1 if x in not_cross_sell_account_id else 0)\n",
        "        df_val['label_not_cross_sell']=df_val['account_id'].apply(lambda x: 1 if x in not_cross_sell_account_id else 0)\n",
        "      if is_segment:\n",
        "        df_train,df_val=segmentation(df_train,df_val,k=4)\n",
        "      if is_WoE:\n",
        "        df_train,df_val=WoE(df_train,df_val,columns=['airlines_name'],target='is_cross_sell',adjusted=adjusted_WoE)\n",
        "      if is_mean_encode:\n",
        "        df_train,df_val=mean_encoding(df_train,df_val,columns=['account_id'],target='is_cross_sell',adjusted=adjusted_mean)\n",
        "      if is_binning:\n",
        "        df_train,df_val=detailed_segmentation(df_train,df_val,columns=['member_duration_days','price'],target='is_cross_sell',num_segments=5)\n",
        "      if 'account_id' in df_train.columns:\n",
        "        df_train,df_val=df_train.drop('account_id',1),df_val.drop('account_id',1)\n",
        "      if 'order_id' in df_train.columns:\n",
        "        df_train,df_val=df_train.drop('order_id',1),df_val.drop('order_id',1)\n",
        "      print('Done Feature Engineering!')\n",
        "    else:\n",
        "      df_train,df_val=df_train.drop('account_id',1),df_val.drop('account_id',1)\n",
        "\n",
        "    df_train_h2o = h2o.H2OFrame(df_train)\n",
        "    df_val_h2o = h2o.H2OFrame(df_val)\n",
        "    #specify the target and predictor variable\n",
        "    x=df_train.drop('is_cross_sell',1).columns.tolist()\n",
        "    y='is_cross_sell'\n",
        "\n",
        "    #Initiate Model\n",
        "    # model_GBM = H2OXGBoostEstimator(**params)\n",
        "    if(model1=='DRF'):\n",
        "      model_1 = H2ORandomForestEstimator(**params_DRF)\n",
        "    elif(model1=='GBM'):\n",
        "      model_1 = H2OGradientBoostingEstimator(**params_GBM)\n",
        "    else :\n",
        "      raise ValueError('Model tidak terdaftar')\n",
        "\n",
        "    if(model2=='DRF'):\n",
        "      model_2 = H2ORandomForestEstimator(**params_DRF)\n",
        "    elif(model2=='GBM'):\n",
        "      model_2 = H2OGradientBoostingEstimator(**params_GBM)\n",
        "    else :\n",
        "      raise ValueError('Model tidak terdaftar') \n",
        "    list_models=[model_1, model_2]\n",
        "    \n",
        "    model_1.train(x = x, y = y, training_frame = df_train_h2o,validation_frame = df_val_h2o)\n",
        "    model_2.train(x = x, y = y, training_frame = df_train_h2o,validation_frame = df_val_h2o)\n",
        "    \n",
        "    ensemble = H2OStackedEnsembleEstimator(base_models=list_models)\n",
        "    ensemble.train(x = x, y = y, training_frame = df_train_h2o,validation_frame = df_val_h2o)\n",
        "\n",
        "    prediction = ensemble.predict(df_val_h2o)[:,2]\n",
        "    prediction_prob=prediction.as_data_frame()['Yes']\n",
        "    for k in f1_threshold :\n",
        "        prediction=prediction_prob.apply(lambda x: \"Yes\" if x>k else \"No\")\n",
        "        # calculate the new F1\n",
        "        precision=confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]/(confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]+confusion_matrix(df_val['is_cross_sell'],prediction)[0][1])\n",
        "        recall=confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]/(confusion_matrix(df_val['is_cross_sell'],prediction)[1][1]+confusion_matrix(df_val['is_cross_sell'],prediction)[1][0])\n",
        "        f1=2*precision*recall/(precision+recall)\n",
        "#         print('Threshold : {}, F1 Score: {}'.format(k,f1))\n",
        "        f1_list.append(f1)\n",
        "  hasil = pd.DataFrame(np.array(f1_list).reshape(5,len(f1_threshold)).transpose())\n",
        "  hasil.index = f1_threshold\n",
        "  hasil['mean'] = hasil.mean(axis=1)\n",
        "  hasil['std']=hasil.std(axis=1)\n",
        "  return hasil"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "MZvxkkfKt6W1"
      },
      "source": [
        "## DRF Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cHqZx-YpHE-Q",
        "colab_type": "text"
      },
      "source": [
        "Finding the best threshold so we got the highest F1-Score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "tSPVLMlrnvhU",
        "outputId": "8c99eb36-67c0-4d91-864e-aaada2ec9cd3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "tune_df=cross_val(df_raw,drop_columns=[x for x in drop_columns if (x!='account_id') & (x!='is_cross_sell') & (x!='order_id') & (x!='cnt_trx') & (x!='airlines_name')],\n",
        "                  f1_threshold=np.arange(0.0,0.45,0.005),params=params_DRF,nfolds=5,model='drf',is_feat_eng=True,is_binning=False,is_WoE=True,adjusted_WoE=True,\n",
        "                  is_mean_encode=True,adjusted_mean=False,is_segment=False,is_PMI=False,is_embedding=False)\n",
        "tune_df.sort_values(by='mean',ascending=False)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "drf prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "drf prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "drf prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "drf prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "drf prediction progress: |████████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "      <th>4</th>\n",
              "      <th>mean</th>\n",
              "      <th>std</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.285</th>\n",
              "      <td>0.435625</td>\n",
              "      <td>0.414667</td>\n",
              "      <td>0.426740</td>\n",
              "      <td>0.425866</td>\n",
              "      <td>0.431539</td>\n",
              "      <td>0.426888</td>\n",
              "      <td>0.007048</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.270</th>\n",
              "      <td>0.434902</td>\n",
              "      <td>0.413483</td>\n",
              "      <td>0.425976</td>\n",
              "      <td>0.426740</td>\n",
              "      <td>0.431978</td>\n",
              "      <td>0.426616</td>\n",
              "      <td>0.007353</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.295</th>\n",
              "      <td>0.434863</td>\n",
              "      <td>0.415490</td>\n",
              "      <td>0.427650</td>\n",
              "      <td>0.425453</td>\n",
              "      <td>0.429180</td>\n",
              "      <td>0.426527</td>\n",
              "      <td>0.006337</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.280</th>\n",
              "      <td>0.435343</td>\n",
              "      <td>0.414447</td>\n",
              "      <td>0.425765</td>\n",
              "      <td>0.425610</td>\n",
              "      <td>0.431465</td>\n",
              "      <td>0.426526</td>\n",
              "      <td>0.007063</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.290</th>\n",
              "      <td>0.434381</td>\n",
              "      <td>0.414169</td>\n",
              "      <td>0.426997</td>\n",
              "      <td>0.426123</td>\n",
              "      <td>0.430667</td>\n",
              "      <td>0.426467</td>\n",
              "      <td>0.006810</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.020</th>\n",
              "      <td>0.161999</td>\n",
              "      <td>0.175869</td>\n",
              "      <td>0.173567</td>\n",
              "      <td>0.166653</td>\n",
              "      <td>0.164222</td>\n",
              "      <td>0.168462</td>\n",
              "      <td>0.005366</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.015</th>\n",
              "      <td>0.162457</td>\n",
              "      <td>0.175735</td>\n",
              "      <td>0.173409</td>\n",
              "      <td>0.166444</td>\n",
              "      <td>0.164147</td>\n",
              "      <td>0.168438</td>\n",
              "      <td>0.005218</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.010</th>\n",
              "      <td>0.160664</td>\n",
              "      <td>0.174537</td>\n",
              "      <td>0.172941</td>\n",
              "      <td>0.164846</td>\n",
              "      <td>0.163939</td>\n",
              "      <td>0.167385</td>\n",
              "      <td>0.005395</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.005</th>\n",
              "      <td>0.160202</td>\n",
              "      <td>0.173884</td>\n",
              "      <td>0.172194</td>\n",
              "      <td>0.163232</td>\n",
              "      <td>0.163736</td>\n",
              "      <td>0.166650</td>\n",
              "      <td>0.005382</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.000</th>\n",
              "      <td>0.154413</td>\n",
              "      <td>0.163811</td>\n",
              "      <td>0.163088</td>\n",
              "      <td>0.150867</td>\n",
              "      <td>0.151729</td>\n",
              "      <td>0.156782</td>\n",
              "      <td>0.005573</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>90 rows × 7 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "              0         1         2         3         4      mean       std\n",
              "0.285  0.435625  0.414667  0.426740  0.425866  0.431539  0.426888  0.007048\n",
              "0.270  0.434902  0.413483  0.425976  0.426740  0.431978  0.426616  0.007353\n",
              "0.295  0.434863  0.415490  0.427650  0.425453  0.429180  0.426527  0.006337\n",
              "0.280  0.435343  0.414447  0.425765  0.425610  0.431465  0.426526  0.007063\n",
              "0.290  0.434381  0.414169  0.426997  0.426123  0.430667  0.426467  0.006810\n",
              "...         ...       ...       ...       ...       ...       ...       ...\n",
              "0.020  0.161999  0.175869  0.173567  0.166653  0.164222  0.168462  0.005366\n",
              "0.015  0.162457  0.175735  0.173409  0.166444  0.164147  0.168438  0.005218\n",
              "0.010  0.160664  0.174537  0.172941  0.164846  0.163939  0.167385  0.005395\n",
              "0.005  0.160202  0.173884  0.172194  0.163232  0.163736  0.166650  0.005382\n",
              "0.000  0.154413  0.163811  0.163088  0.150867  0.151729  0.156782  0.005573\n",
              "\n",
              "[90 rows x 7 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 56
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wXIiMka_HE-T",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "best_f1_DRF=tune_df.sort_values(by='mean',ascending=False).index[0]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VYFIm7IJHE-X",
        "colab_type": "text"
      },
      "source": [
        "# GBM Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SIhPX8lvHE-Y",
        "colab_type": "text"
      },
      "source": [
        "Finding the best threshold so we got the highest F1-Score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pCL_KTbXHE-Z",
        "colab_type": "code",
        "outputId": "dc0e056a-e532-4a5f-bb4a-f823458567ad",
        "colab": {}
      },
      "source": [
        "tune_df=cross_val(df_raw,drop_columns=[x for x in drop_columns if (x!='account_id') & (x!='is_cross_sell') & (x!='order_id') & (x!='cnt_trx') & (x!='airlines_name')],\n",
        "                  f1_threshold=np.arange(0.0,0.45,0.005),params=params_GBM,nfolds=5,model='gbm',is_feat_eng=True,is_binning=False,is_WoE=True,adjusted_WoE=True,\n",
        "                  is_mean_encode=True,adjusted_mean=False,is_segment=False,is_PMI=False,is_embedding=False)\n",
        "tune_df.sort_values(by='mean',ascending=False)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "gbm prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "gbm prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "gbm prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "gbm prediction progress: |████████████████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAA/YAAAHVCAYAAABbv2DlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3XucXWV97/HPl4QGuRgUkAZrG6oRBIIgAwgGGpTaapSCYPFyFJAatSreqOZoVRStofZ1xEstjRRBoegRhaJpBVTCVZQJ5AIIeJR4KuIdU64Rkt/5Y6/UfYa57ElmMlmZz/v1mtfstdaznue31sw/3/08a+9UFZIkSZIkqZ22mugCJEmSJEnShjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFps60QVIg9l5551r5syZE12GJEmSJE2IpUuX/rKqdumlrcFem6WZM2fS398/0WVIkiRJ0oRI8qNe27oUX5IkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS3m191ps7Ty7tXMXLB4osuQJEmStAVbtXDeRJcwJpyxlyRJkiSpxQz2kiRJkiS1mMFekiRJkqQWM9hLkiRJktRiBntJkiRJklpsUgT7JDOTvGKENn1JPjHEsVVJdh6f6tonydwkX2teH5VkwUTXJEmSJEmT1WT5uruZwCuAfx2qQVX1A/2bqqAtRVVdClw60XVIkiRJ0mQ1rjP2SS5JsjTJrUnmN/v+PMlNSZYn+Wazb/skn02yMsmKJMc2+1/e7LslyRld/d7f9fq4JOc2r89N8okk1yf5YZLjmmYLgcOSLEvytiFq7Z6F3inJ5UluTvLPQIa5xplJbk9ydlPnBUmOTHJdku8nOahpt12Sc5Lc2PT7F13nX9Pck5uSHNpVz5IkFzX9X5BkuDoOSHJVc78vSzKj2b8kyRlJvpvkziSHNfunJPmHrnv+5mb/85r6Vjb1Tuv6u92e5FrgJV3jnpjkU8Pd/yRbJfl083/wtST/3vW3kSRJkiRthPFeiv+aqjoA6ANOSbIr8Bng2Kp6JvDSpt17gdVVNbuq9gW+lWQ34AzgucB+wIFJju5hzBnAHOBFdAI9wALgmqrar6o+1kMf7weurar96cxG/+EI7Z8GfBzYF9iTzuqAOcCpwLubNu8BvlVVBwJHAB9Nsh3wc+BPq+pZwPFA9+MA+wNvBfYC/hh4zmCDJ9ka+CRwXHO/zwE+3NVkalUd1PT1/mbffGB3YP/mnl+QZBvgXOD4qppNZ0XHG5r9nwFeDBwG/P4w92Kw+/8SOqsmZgN/BRwyxHXMT9KfpH/tg6uHGUKSJEmStN54B/tTkiwHbgCeQidMXl1VdwFU1a+bdkcC/7j+pKq6FzgQWFJVv6iqR4ELgMN7GPOSqlpXVbcBu25g3YcD5ze1LAbuHaH9XVW1sqrWAbcC36yqAlbSCbQAzwcWJFkGLAG2ofOGwdbAZ5KsBL5EJ8Sv992q+nHT77KuvgbaA9gHuKLp/2+BP+g6/pXm99KuPo4Ezmru7fq/xR7NtdzZtDmvuRd7Nvu/31zX+cPci8Hu/xzgS83+nwJXDnZiVS2qqr6q6puy7fRhhpAkSZIkrTduz9gnmUsnPB5SVQ8mWQIspxMeH9McqEH2DaW77TYDjq3psY+RDKxnON1jruvaXsfv7nHorFS4o/vEJKcBPwOeSeeNloeH6HctQ/+9AtxaVYPOhHf1093Hxtzz4Qx2/zfm7yBJkiRJGsZ4zthPB+5tQv2ewLOBacCfJNkdIMkTm7aXA29af2KSJwDfadrunGQK8HLgqqbJz5I8I8lWwDE91HIfsMMoar8aeGVTywuAJ4zi3KFcBrx5/XPySfZv9k8H7mlm5V8FTNmAvu8AdklySNP31kn2HuGcy4HXJ5nanPNE4HZgZpKnNW1eReee3w7snuSpzf6Xj7K+a4Fjm2ftdwXmjvJ8SZIkSdIQxjPYfx2YmmQFcDqd5fi/oLMc/yvNEv0vNm0/BDyh+fC55cARVXUP8D/pLNteDtxUVf/WtF8AfA34FnBPD7WsAB5N5wP7Bv3wvAE+ABye5CY6S+j/bw/njOR0OsvuVyS5pdkG+DRwQpIbgKcDD4y246r6LXAccEZz/5YBh45w2tl0rmtFc84rquph4CTgS82jAevoLNd/mM7fbXHz4Xk/GmWJXwZ+DNwC/DOdN218iF6SJEmSxkA6j0xL4yvJ9lV1f5KdgO8Cz2metx/UtBmzasYJZ266AiVJkiRNOqsWzpvoEoaUZGlV9fXSdrJ8j70m3teS7Aj8HnD6cKFekiRJktS7SRfsk/wZna/R63ZXVQ37rH4z0/zNQQ49r6p+NVb1jSTJxXS+pq7bu6rqsk1Vw4aoqrkTXYMkSZIkbYlciq/NUl9fX/X39090GZIkSZI0IUazFH+8v8dekiRJkiSNI4O9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSi02d6AKkway8ezUzFyzeZOOtWjhvk40lSZIkSWPJGXtJkiRJklrMYC9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWaw15hIcmKS3Sa6DkmSJEmabAz2GisnAoMG+yRTNm0pkiRJkjR5GOwnmSSXJFma5NYk85O8Icnfdx0/Mcknm9fvTXJ7kiuSXJjk1CH6PA7oAy5IsizJ45KsSvK+JNcCL09yY5K5TfuPJPnw+F+tJEmSJG35/B77yec1VfXrJI8DbgSeB1wHvLM5fjzw4SR9wLHA/nT+T24Clg7WYVVdlORNwKlV1Q+QBODhqprTbN8MXJTkFODPgYMH9pNkPjAfYMrjdxmbq5UkSZKkLZwz9pPPKUmWAzcATwF2B36Y5NlJdgL2oBP05wD/VlUPVdV9wFc3YKwvrn9RVbcCn2/6eU1V/XZg46paVFV9VdU3ZdvpGzCcJEmSJE0+zthPIs1S+COBQ6rqwSRLgG3oBPC/BG4HLq6qSjPlvpEeGLA9G/gNsOsY9C1JkiRJwhn7yWY6cG8T6vcEnt3s/wpwNPByfjfLfi3w4iTbJNkemDdC3/cBOwx1MMlLgJ2Aw4FPJNlxwy9DkiRJkrSewX5y+TowNckK4HQ6y/GpqnuB24A/qqrvNvtuBC4FltMJ/v3A6mH6Phc4a/2H53UfSLIzsBA4uaruBD4FfHwMr0uSJEmSJq1U1UTXoM1Uku2r6v4k2wJXA/Or6qZNMfa0GbNqxglnboqhAFi1cKQFCZIkSZK06SRZWlV9vbT1GXsNZ1GSveg8h3/epgr1kiRJkqTeGew1pKp6xcB9Sf4ReM6A3R+vqs9umqokSZIkSd1ciq/NUl9fX/X39090GZIkSZI0IUazFN8Pz5MkSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS02daILkAaz8u7VzFyweMz7XbVw3pj3KUmSJEkTyRl7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9gMkWZKkb5zHODfJcWPU18wkr+ja7kvyibHoe4jxTkty6nj1L0mSJEkaHYP9GEoyLl8fOEK/M4H/DvZV1V9Vp4xHHZIkSZKkzU+rgn0zO317krOT3JLkgiRHJrkuyfeTHJRkuyTnJLkxyc1J/qI598QklyT5apK7krwpydubNjckeWLXUP8jyfXNGAc15w/X75eSfBW4fIi6k+RTSW5Lshh4UtexVUl2bl73JVnSvD4tyaIklwOfa679miQ3NT+HNl0sBA5LsizJ25LMTfK1po8nNte8ornGfbv6PqdZnfDDJMO+EZDkPUnuSPINYI+u/a9t7sfyJF9Osm2SHZr7u3XT5vHNNW6d5JTmHqxI8oXe/uqSJEmSpOGMywzzOHsa8FJgPnAjndnqOcBRwLuB24BvVdVrkuwIfLcJpAD7APsD2wD/B3hXVe2f5GPAq4Ezm3bbVdWhSQ4HzmnOe88w/R4C7FtVvx6i5mPoBOLZwK5Njef0cK0HAHOq6qEk2wJ/WlUPJ5kFXAj0AQuAU6vqRQBJ5nad/wHg5qo6Oslzgc8B+zXH9gSOAHYA7kjyT1X1yMACkhwAvKy5b1OBm4ClzeGvVNVnmnYfAk6uqk82b07MAy5pzv1yVT2SZAGwe1Wtae7hwLHm0/m7MuXxu/RweyRJkiRJbQz2d1XVSoAktwLfrKpKspLOsvQ/AI7qeg58G+APm9dXVtV9wH1JVgNfbfavBPbtGuNCgKq6uplx3hF4/jD9XjFMqAc4HLiwqtYCP0nyrR6v9dKqeqh5vTXwqST7AWuBp/dw/hzg2OZavpVkpyTTm2OLq2oNsCbJz+m84fDjQfo4DLi4qh4ESHJp17F9mkC/I7A9cFmz/2zgnXSC/UnAa5v9K4ALklzSHPv/VNUiYBHAtBmzqofrkyRJkqRJr43Bfk3X63Vd2+voXM9a4NiquqP7pCQH93DuegNDZQEZpt8Heqh7qKD6KL97JGKbAce6+30b8DPgmU37h3sYM8PU0X0v1jL8/8JQtZ8LHF1Vy5OcCMwFqKrrmkcH/gSYUlW3NO3n0XmT4yjgvUn2rqpHe7gOSZIkSdIQWvWMfY8uA96cJABJ9t+APo5vzp0DrK6q1RvZ79XAy5JMSTKDzhL49VbRWXIPzez6EKYD91TVOuBVwJRm/310ltMPNe4rm3rnAr+sqv8aRd3r+zgmyeOS7AC8uOvYDsA9zfP0rxxw3uforHz4bDP+VsBTqupKOrP562f5JUmSJEkbYUsM9qfTWba+IsktzfZo3ZvkeuAs4OQx6Pdi4Pt0lvz/E3BV17EPAB9Pcg2dmfOhfBo4IckNdJbhr5/NXwE82nyA3dsGnHMa0JdkBZ0P2TthFDUDUFU3AV8ElgFfBq7pOvxe4DvAFcDtA069AHgCzWMNdN6IOL95ZOJm4GNV9ZvR1iNJkiRJ+v+lykeZNfaSHAf8RVW9akPOnzZjVs044cyRG47SqoXzxrxPSZIkSRprSZZWVV8vbdv4jL02c0k+CbwAeOFE1yJJkiRJWzqD/RhKMhv4/IDda6rq4ImoZzSS7AR8c5BDz6uqX42mr6p689hUJUmSJEkaiUvxtVnq6+ur/v7+iS5DkiRJkibEaJbib4kfnidJkiRJ0qRhsJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWqxqRNdgDSYlXevZuaCxWPa56qF88a0P0mSJEnaHDhjL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV7/Lclbk2w7xLETk3xqmHOPTrLX+FUnSZIkSRqMwV7d3goMGux7cDQwaLBP4tcqSpIkSdI4MdhPUkm2S7I4yfIktyR5P7AbcGWSK5s2JyW5M8lVwHOG6etQ4Cjgo0mWJXlqkiVJ/q459y1J/i3Jq5v2r0tywSD9zE/Sn6R/7YOrx+OyJUmSJGmL40zq5PXnwE+qah5AkunAScARVfXLJDOADwAHAKuBK4GbB+uoqq5Pcinwtaq6qOkPYMeq+pNm+3zguiR3Ae8Anj1IP4uARQDTZsyqMbxWSZIkSdpiOWM/ea0EjkxyRpLDqmrgFPnBwJKq+kVV/Rb44gaM8d/nVNXPgPfReYPgHVX16w0tXJIkSZL0O87YT1JVdWeSA4AXAh9JcvlgzTZymAcGbM8GfkVnyb8kSZIkaQw4Yz9JJdkNeLCqzgf+AXgWcB+wQ9PkO8DcJDsl2Rp46Qhddp872HgHAS8A9gdOTbL7Rl6CJEmSJAln7Cez2XQ+7G4d8AjwBuAQ4D+S3FNVRyQ5Dfg2cA9wEzBlmP6+AHwmySnAcd0HkkwDPgOcVFU/SfIO4Jwkz60qn6WXJEmSpI0Qc5U2R9NmzKoZJ5w5pn2uWjhvTPuTJEmSpPGSZGlV9fXS1qX4kiRJkiS1mEvxNSpJ3sNjn7f/UlV9eCzHmf3k6fQ7wy5JkiRJIzLYa1SaAD+mIV6SJEmStOFcii9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxQz2kiRJkiS1mMFekiRJkqQWM9hLkiRJktRiBntJkiRJklrMYC9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxQz2kiRJkiS1mMFekiRJkqQWM9hLkiRJktRiUye6AGkwK+9ezcwFi8ekr1UL541JP5IkSZK0OXLGXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg/0kk+TEJJ9qXr8+yavHoM++JJ/Y+OokSZIkSaPl191NYlV11sb2kWRqVfUD/WNQkiRJkiRplJyx38IkuSTJ0iS3Jpnf7DspyZ1JrgKe09X2tCSnDtPXkiRnJrk+yS1JDuo6b1GSy4HPJZmb5GvNse2TfDbJyiQrkhzb7H9+km8nuSnJl5JsP8h485P0J+lf++Dqsb0xkiRJkrSFcsZ+y/Oaqvp1kscBNyZZDHwAOABYDVwJ3DyK/rarqkOTHA6cA+zT7D8AmFNVDyWZ29X+vcDqqpoNkOQJSXYG/hY4sqoeSPIu4O3AB7sHqqpFwCKAaTNm1aiuWpIkSZImKYP9lueUJMc0r58CvApYUlW/AEjyReDpo+jvQoCqujrJ45Ps2Oy/tKoeGqT9kcDL1m9U1b1JXgTsBVyXBOD3gG+PogZJkiRJ0hAM9luQZub8SOCQqnowyRLgduAZG9HtwJnz9dsPDFXGIOcEuKKqXr4RdUiSJEmSBuEz9luW6cC9TajfE3g28DhgbpKdkmwNvHSUfR4PkGQOnSX2Iz38fjnwpvUbSZ4A3AA8J8nTmn3bJhnNqgFJkiRJ0hAM9luWrwNTk6wATqcTqO8BTqOz9P0bwE0DzhnpWfZ7k1wPnAWc3EMNHwKe0HzY3nLgiOYxgBOBC5vabgD27OmKJEmSJEnDcin+FqSq1gAvGOTQEuCzg+zfCfjRCN1+uar+54BxThuwvaQZg6q6HzhhkNq+BRw4wliSJEmSpFFyxn6SSnI6cDBw6UTXIkmSJEnacM7YT1JV9V46X01Hkn+k6/vtGx+vqrmbui5JkiRJ0uikyq8L1+anr6+v+vv7J7oMSZIkSZoQSZZWVV8vbV2KL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWmzqRBcgDWbl3auZuWDxmPS1auG8MelHkiRJkjZHzthLkiRJktRiBntJkiRJklrMYC9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawn+SSfDDJkePU97lJjmten51kr+b1S5N8L8mV4zGuJEmSJE0mft3dJJZkSlW9b1OMVVV/1bV5MvDXVWWwlyRJkqSN5Iz9FirJzCS3JzkvyYokFyXZNsmqJO9Lci3w0gGz6gcmuT7J8iTfTbJDkilJPprkxqaf1w0zZpJ8KsltSRYDT+o6tiRJX5L3AXOAs5J8dMD585P0J+lf++Dq8bkxkiRJkrSFMdhv2fYAFlXVvsB/AX/d7H+4quZU1RfWN0zye8AXgbdU1TOBI4GH6Myur66qA4EDgdcm2X2I8Y5pxpwNvBY4dGCDqvog0A+8sqr+ZsCxRVXVV1V9U7advsEXLUmSJEmTicF+y/afVXVd8/p8OjPl0AnwA+0B3FNVNwJU1X9V1aPA84FXJ1kGfAfYCZg1xHiHAxdW1dqq+gnwrTG6DkmSJEnSEHzGfstWQ2w/MEjbDNJ+/f43V9VlGzimJEmSJGkcOWO/ZfvDJIc0r18OXDtM29uB3ZIcCNA8Xz8VuAx4Q5Ktm/1PT7LdEH1cDbyseS5/BnDEmFyFJEmSJGlIBvst2/eAE5KsAJ4I/NNQDavqt8DxwCeTLAeuALYBzgZuA25Kcgvwzwy90uNi4PvAymasq8boOiRJkiRJQ0iVK6e3RElmAl+rqn0muJQNMm3GrJpxwplj0teqhfPGpB9JkiRJ2lSSLK2qvl7aOmMvSZIkSVKL+eF5W6iqWgWMy2x9ktnA5wfsXlNVB4/VGLOfPJ1+Z9olSZIkaUQGe41aVa0E9pvoOiRJkiRJLsWXJEmSJKnVDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSi40Y7JPsmuRfkvxHs71XkpPHvzRJkiRJkjSSXmbszwUuA3Zrtu8E3jpeBUmSJEmSpN71Eux3rqr/DawDqKpHgbXjWpUkSZIkSepJL8H+gSQ7AQWQ5NnA6nGtSpIkSZIk9WRqD23eDlwKPDXJdcAuwHHjWpUkSZIkSerJiMG+qm5K8ifAHkCAO6rqkXGvTJIkSZIkjWjIYJ/kJUMcenoSquor41STJEmSJEnq0XAz9i8e5lgBBntJkiRJkibYkMG+qk7alIVIkiRJkqTRG/EZ++YT8d8PzKEzU38t8MGq+tU416ZJbOXdq5m5YPGY9LVq4bwx6UeSJEmSNke9fN3dF4BfAMfS+TT8XwBfHM+iJEmSJElSb3r5ursnVtXpXdsfSnL0eBUkSZIkSZJ618uM/ZVJXpZkq+bnL4GxWSMtSZIkSZI2ynBfd3cfnWfqA7wdOL85tBVwP53n7iVJkiRJ0gQa7lPxd9iUhUiSJEmSpNHr5Rl7kjwBmAVss35fVV09XkVp4iQ5CtirqhaO8ryZwKFV9a/jUZckSZIkaXC9fN3dXwFvAf4AWAY8G/g28NzxLU0ToaouBS7dgFNnAq8AHhPsk0ytqkc3sjRJkiRJ0iB6+fC8twAHAj+qqiOA/el85Z1aJsnMJLcnOTvJLUkuSHJkkuuSfD/JQUlOTPKppv25ST6R5PokP0xy3DDdLwQOS7Isyduafr6U5KvA5UmOSfKNdMxIcmeS3x9Q3/wk/Un61z64ehzvhCRJkiRtOXoJ9g9X1cMASaZV1e3AHuNblsbR04CPA/sCe9KZZZ8DnAq8e5D2M5rjL6IT3oeyALimqvarqo81+w4BTqiq51bVxcBPgTcCnwHeX1U/7e6gqhZVVV9V9U3ZdvoGX6AkSZIkTSa9PGP/4yQ7ApcAVyS5F/jJ+JalcXRXVa0ESHIr8M2qqiQr6SynH+iSqloH3JZk11GOdUVV/bpr+83ALcANVXXhBtQuSZIkSRpgxGBfVcc0L09LciUwHfj6uFal8bSm6/W6ru11DP7/0N0+oxzrgQHbT27G2TXJVs0bBpIkSZKkjTDkUvwkj29+P3H9D7ASuBbYfhPVp/a4DxjyKxKTTAU+S2fp//eAt2+iuiRJkiRpizbcjP2/0nmueilQdGZru3//8bhXpzZZATyaZDlwLnDvgOPvpvMM/jVJlgE3JllcVd/bxHVKkiRJ0hYlVTX0wSTAU6rq/266kiSYNmNWzTjhzDHpa9XCeWPSjyRJkiRtKkmWVlVfL22H/VT86qT+i8ekKkmSJEmSNOZ6+VT8G5IcWFU3jns12uwlmQ18fsDuNVV18ETUI0mSJEmT3bBL8QGS3AY8HfgRnU85D53J/H3HvzxNVn19fdXf3z/RZUiSJEnShBjNUvxeZuxfsJH1SJIkSZKkcdLL99j/CCDJk4Btxr0iSZIkSZLUs2E/PA8gyVFJvg/cBVwFrAL+Y5zrkiRJkiRJPRgx2AOnA88G7qyq3YHnAdeNa1WSJEmSJKknvQT7R6rqV8BWSbaqqiuB/ca5LkmSJEmS1INePjzvN0m2B64BLkjyc+DR8S1LkiRJkiT1opcZ+6uBHYG3AF8HfgC8eDyLkiRJkiRJvekl2Ae4DFgCbA98sVmaL0mSJEmSJtiIwb6qPlBVewNvBHYDrkryjXGvTJIkSZIkjaiXGfv1fg78FPgV8KTxKUeSJEmSJI1GL99j/4YkS4BvAjsDr62qfce7MEmSJEmSNLJePhX/j4C3VtWy8S5GkiRJkiSNzojBvqoWbIpCJEmSJEnS6I3mGXtJkiRJkrSZMdhLkiRJktRivTxjL21yK+9ezcwFi8ekr1UL541JP5IkSZK0OXLGXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGew1rCQzk7xihDb7JXnhpqpJkiRJkvQ7BnuNZCYwbLAH9gMGDfZJ/OYFSZIkSRpHhq5JKsmrgVOBAlYAa4H/AvqA3wfeWVUXAQuBZyRZBpxXVR8b0M/vAR8EHpdkDvAR4BnAbnTeFPhlkn5gn6p6TZLZwIXAQVX14PhfqSRJkiRt2Qz2k1CSvYH3AM+pql8meSLwv4AZwBxgT+BS4CJgAXBqVb1osL6q6rdJ3gf0VdWbmv5PAw4A5lTVQ0m2ApYkOaYZ93WGekmSJEkaGy7Fn5yeC1xUVb8EqKpfN/svqap1VXUbsOtGjnFpVT3U9L8OOBH4PHBVVV032AlJ5ifpT9K/9sHVGzm8JEmSJE0OBvvJKXSW4A+0ZkCbjfHAgO1ZwP10lugPqqoWVVVfVfVN2Xb6Rg4vSZIkSZODwX5y+ibwl0l2AmiW4g/lPmCHEfobtk2S6cDHgcOBnZIcN7pyJUmSJEmhN1PwAAAW0klEQVRDMdhPQlV1K/Bh4Koky+k8Xz+UFcCjSZYnedsQba4E9kqyLMnxgxz/GPDpqroTOBlYmORJG3EJkiRJkqSGH543SVXVecB5wxzfvvn9CPC8Efr6NXDgMMdf0/X6P4GnjbZeSZIkSdLgnLGXJEmSJKnFnLFXz5L8GXDGgN13VdUxE1GPJEmSJMlgr1GoqsuAyya6DkmSJEnS7xjstVma/eTp9C+cN9FlSJIkSdJmz2fsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSi02d6AKkway8ezUzFyzuuf2qhfPGsRpJkiRJ2nw5Yy9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxQz2GhNJ5iY5dKLrkCRJkqTJxmCvsTIXGDTYJ/HbFyRJkiRpnBjst3BJZia5PcnZSW5JckGSI5Ncl+T7SQ5qfq5PcnPze4/m3LcnOad5Pbs5f9vBxgBeD7wtybIkhyU5N8n/SnIlcEaSTyR5X9P+z5JcncT/P0mSJEnaSM6kTg5PA14KzAduBF4BzAGOAt4NvBo4vKoeTXIk8HfAscCZwJIkxwDvAV5XVQ8O7LyqViU5C7i/qv4BIMnJwNOBI6tqbfOGwI1JrgE+AbywqtZ195NkflMjUx6/y1jfA0mSJEnaIhnsJ4e7qmolQJJbgW9WVSVZCcwEpgPnJZkFFLA1QFWtS3IisAL456q6bpTjfqmq1jZ9PZjktcDVwNuq6gcDG1fVImARwLQZs2r0lylJkiRJk49LoSeHNV2v13Vtr6Pz5s7pwJVVtQ/wYmCbrvazgPuB3TZg3AcGbM8GfrWBfUmSJEmSBmGwF3Rm7O9uXp+4fmeS6cDHgcOBnZIcN0wf9wE7DHUwyR8B7wD2B16Q5OCNrFmSJEmShMFeHX8PfCTJdcCUrv0fAz5dVXcCJwMLkzxpiD6+Chyz/sPzug8kCfAvwKlV9ZOmr7OTbDNIP5IkSZKkUUiVjzJr8zNtxqyaccKZPbdftXDeOFYjSZIkSZtWkqVV1ddLW2fsJUmSJElqMT8VX6OS5CTgLQN2X1dVb5yIeiRJkiRpsjPYa1Sq6rPAZye6DkmSJElSh8Fem6XZT55Ov8/NS5IkSdKIfMZekiRJkqQWM9hLkiRJktRiBntJkiRJklrMYC9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxQz2kiRJkiS1mMFekiRJkqQWM9hLkiRJktRiBntJkiRJklrMYC9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxaZOdAHSYFbevZqZCxaP2G7VwnmboBpJkiRJ2nw5Yy9JkiRJUosZ7CVJkiRJajGDvSRJkiRJLWawlyRJkiSpxQz2kiRJkiS1mMF+EEn+PcmOQxxblWTn5vX1m7ayiZHk4iRHd23fkeRvu7a/nOQlw5w/N8nqJMu6fo4c77olSZIkaTLw6+4GUVUvHLgvSYAMaHfoJitqYl0PHApckmQn4H7gkK7jhwBvHKGPa6rqReNUnyRJkiRNWpN+xj7JJUmWJrk1yfxm36okOyeZmeR7ST4N3AQ8ZcC59ze/5yZZkuSiJLcnuaB5I4AkByS5qhnjsiQzmv2nJLktyYokXximvtOSnNP0/8MkpwxX+/q6kpzRHPtGkoO6zj+qaTMlyUeT3NjU8LphbtN1dII9ze+vAbukY3fgoar6aZJtknw2ycokNyc5ote/gyRJkiRpwzhjD6+pql8neRxwY5IvDzi+B3BSVf01QJPXB7M/sDfwEzpB+DlJvgN8EviLqvpFkuOBDwOvARYAu1fVmqGW/XfZEzgC2AG4I8k/VdUjg9VeVb8CtgOWVNW7klwMfAj4U2Av4DzgUuBkYHVVHZhkGnBdksur6q5Bxl8K7JPk9+gE+6uAPwae0Vz3dU27NwJU1ewkewKXJ3l6c+ywJMu6+jy2qn7QPUjz5sR8gCmP32WEWyJJkiRJAoM9wClJjmlePwWYNeD4j6rqhh76+W5V/RigCbAzgd8A+wBXNG8ITAHuadqvAC5IcglwyQh9L66qNcCaJD8HdgV+PETtvwJ+C3y92b8SWFNVjyRZ2dQF8Hxg3yTHNdvTm/MfE+ybNx9uBZ4FPBv4ezrB/lA6wX79Zw3MofNGBlV1e5IfAeuD/YhL8atqEbAIYNqMWTXCPZEkSZIkMcmDfZK5wJHAIVX1YJIlwDYDmj3QY3drul6vpXNvA9xaVYcM0n4ecDhwFPDeJHtX1aO99j1C7Y9U1fpgvG79+VW1Lsn6v3mAN1fVZT1e3/VNvTtU1b1JbgDeRCfYn9XVpyRJkiRpE5rsz9hPB+5tgvGedGajx9IddJ5FPwQgydZJ9k6yFfCUqroSeCewI7D9KPve2NovA96QZOumtqcn2W6Y9tcBrwOWN9srmjH/ELi12Xc18Mr1/TXH7hhlXZIkSZKkUZjUM/Z0lqu/PskKOgG0lyX3Pauq3zZL3T+RZDqd+30mcCdwfrMvwMeq6jej7H5jaz+bzrL8m5oP+vsFcPQw7a+ns/z+IwBV9WjzWMB/VtW6ps2ngbOaJf+PAic2y/jhsc/Yf6iqLhplzZIkSZKkAfK7FdvS5mPajFk144QzR2y3auG8TVCNJEmSJG1aSZZWVV8vbSf7UnxJkiRJklptsi/F32wkOQl4y4Dd11XVGzdhDbOBzw/YvaaqDt5UNUiSJEmSRsel+Nos9fX1VX9//0SXIUmSJEkTwqX4kiRJkiRNEgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgZ7SZIkSZJazGAvSZIkSVKLGewlSZIkSWoxg70kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFps60QVIg1l592pmLlj8mP2rFs6bgGokSZIkafPljL0kSZIkSS1msJckSZIkqcUM9pIkSZIktZjBXpIkSZKkFjPYS5IkSZLUYgb7zVCS6zfgnHePRy2SJEmSpM2bwX4zVFWHbsBp4xrsk0wZz/4lSZIkSRvGYL8ZSnJ/83tGkquTLEtyS5LDhmi/EHhc0+6CJAcmWZFkmyTbJbk1yT5DnDu3GePiJLclOSvJVuvrSPLBJN8BDknyvCQ3J1mZ5Jwk05p2q5L8XZJvJ+lP8qwklyX5QZLXN22S5KPNdaxMcvwgtcxvzu9f++DqsbmZkiRJkrSFM9hv3l4BXFZV+wHPBJYN1qiqFgAPVdV+VfXKqroRuBT4EPD3wPlVdcsw4xwEvAOYDTwVeEmzfzvglqo6GOgHzgWOr6rZwFTgDV19/GdVHQJc07Q7Dng28MHm+EuA9ddxJPDRJDMGXMeiquqrqr4p204f9sZIkiRJkjoM9pu3G4GTkpwGzK6q+0Zx7geBPwX66IT74Xy3qn5YVWuBC4E5zf61wJeb13sAd1XVnc32ecDhXX1c2vxeCXynqu6rql8ADyfZsenzwqpaW1U/A64CDhzF9UiSJEmSBmGw34xV1dV0wvPdwOeTvHoUpz8R2B7YAdhmpKGG2H64CfsAGaGPNc3vdV2v129P7eF8SZIkSdIGMNhvxpL8EfDzqvoM8C/As4Zp/kiSrbu2FwHvBS4AzhhhqIOS7N48W388cO0gbW4HZiZ5WrP9Kjqz7r26Gjg+yZQku9B5w+K7ozhfkiRJkjSIqRNdgIY1F/ibJI8A9wPDzdgvAlYkuQm4DHi0qv61+TT765M8t6q+NcS53wYW0nnG/mrg4oENqurhJCcBX0oylc5jAmeN4louBg4BltNZEfDOqvrpKM6XJEmSJA0iVQNXYWsySTIXOLWqXjTRtXSbNmNWzTjhzMfsX7Vw3gRUI0mSJEmbVpKlVdXXS1uX4kuSJEmS1GIuxW+Z5jvlpw3Y/aqqWjnCebOBzw/Yvab5KrslY1ehJEmSJGlTcim+Nkt9fX3V398/0WVIkiRJ0oRwKb4kSZIkSZOEwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaSJEmSJLWYwV6SJEmSpBYz2EuSJEmS1GIGe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktNnWiC5AGs/Lu1cxcsPgx+1ctnDcB1UiSJEnS5ssZe0mSJEmSWsxgL0mSJElSixnsJUmSJElqMYO9JEmSJEktZrCXJEmSJKnFDPaTUJI9kyxLcnOSp47jOEcn2Wu8+pckSZIkGewnq6OBf6uq/avqB+M8jsFekiRJksaRwX4zlWRmku8l+UySW5NcnuRxSfZLckOSFUkuTvKEYfp4TNskLwTeCvxVkiuHOG+7JIuTLE9yS5Ljm/0HJLkqydIklyWZ0ex/bZIbm/ZfTrJtkkOBo4CPNqsDnprklCS3NfV8YezvmiRJkiRNPgb7zdss4B+ram/gN8CxwOeAd1XVvsBK4P3DnP+YtlX178BZwMeq6oghzvtz4CdV9cyq2gf4epKtgU8Cx1XVAcA5wIeb9l+pqgOr6pnA94CTq+p64FLgb6pqv2ZlwAJg/6ae1w8cNMn8JP1J+tc+uLrXeyRJkiRJk5rBfvN2V1Uta14vBZ4K7FhVVzX7zgMOH+zEJNN7bTuIlcCRSc5IclhVrQb2APYBrkiyDPhb4A+a9vskuSbJSuCVwN5D9LsCuCDJ/wAeHXiwqhZVVV9V9U3ZdnqPpUqSJEnS5DZ1ogvQsNZ0vV4L7LgpBq2qO5McALwQ+EiSy4GLgVur6pBBTjkXOLqqlic5EZg7RNfz6Ly5cBTw3iR7V9VjAr4kSZIkqXfO2LfLauDeJIc1268CrhqsYTPL3lPbgZLsBjxYVecD/wA8C7gD2CXJIU2brZOsn5nfAbinWa7/yq6u7muOkWQr4ClVdSXwTjpvUmzfSz2SJEmSpKE5Y98+JwBnJdkW+CFw0hi17TabzoferQMeAd7w/9q711DL6jIM4M/bTBnhqOD4QdIaKYsmDY2DdIMKI7qAQkhpSBaSH6SkC4FRUdgXK6IS7KJUVtBVqIas/GBJZWlNGJbBwKBpVmRNJpVdTN8+7F0MM9OZ7cw5e5219+8HA3vt9T/DM/DO3uc567/X6e5/VdU5Sa6YbvPfnOTDSW5P8q4ktyS5K5Nt/Fumf88Xk1xdVZckOTfJJ6dfW5l8xv/Ps/6jAQAAOLDq7qEzwH6OOP7kPv6CD+/3/K8uf/kAaQAAAOarqn7a3SuzrLUVHwAAAEbMVvwFUFVXJnnuPk9/pLs/fZCvOzbJDQc4dWZ371mrfAAAAKwfW/HZkFZWVnrnzp1DxwAAABiErfgAAACwJBR7AAAAGDHFHgAAAEZMsQcAAIARU+wBAABgxBR7AAAAGDHFHgAAAEZMsQcAAIARU+wBAABgxBR7AAAAGDHFHgAAAEZMsQcAAIARU+wBAABgxBR7AAAAGDHFHgAAAEZMsQcAAIARU+zZkH7+m/uz7dLrsu3S64aOAgAAsKEp9gAAADBiij0AAACMmGIPAAAAI6bYAwAAwIgp9hyyqjqmqi5e5fwP55kHAABgGSn2HI5jkuxX7KtqU5J093PmnggAAGDJbB46AKN2eZInVdXPkjyY5K9JfpfktCTbq+qv3X1kVb0gyWVJ9iR5apLvJbm4ux8eJjYAAMDiUOw5HJcmOaW7T5uW9+umx3ceYO0ZSbYnuSvJt5O8Ism1ey+oqouSXJQkm446bh1jAwAALA5b8VlLP/4/pf6/5+7o7oeSfCHJ8/Zd0N1XdfdKd69setzR6xoUAABgUSj2rKW/rXKuD3IMAADAIVDsORx/SbJlxrVnVNVJVfWoJK9K8oP1iwUAALA8fMaeQ9bde6rqpqr6RZK/J/n9Kst/lMnN9k7N5OZ5X51DRAAAgIWn2HNYuvvVq5w7cq/DB7r7VXOIBAAAsFRsxQcAAIARc8WeddfdNya5ceAYAAAAC0mxZ0M69fFHZ+flLx86BgAAwIZnKz4AAACMmGIPAAAAI6bYAwAAwIgp9gAAADBiij0AAACMmGIPAAAAI6bYAwAAwIgp9gAAADBi1d1DZ4D9VNVfkuwaOgdLa2uSPw4dgqVk9hiS+WMoZo8hbeT5e2J3HzfLws3rnQQO0a7uXhk6BMupqnaaP4Zg9hiS+WMoZo8hLcr82YoPAAAAI6bYAwAAwIgp9mxUVw0dgKVm/hiK2WNI5o+hmD2GtBDz5+Z5AAAAMGKu2AMAAMCIKfYAAAAwYoo9g6qql1TVrqraXVWXHuD8EVX1pen5W6pq2/xTsqhmmL+3VNUvq+q2qrqhqp44RE4Wz8Fmb69151RVV9Xofw0PG8Mss1dVr5y+9t1eVZ+fd0YW1wzvu0+oqu9W1a3T996XDZGTxVNVn6qqe6vqF//nfFXVFdPZvK2qnjnvjIdLsWcwVbUpyZVJXppke5Lzqmr7PssuTHJfdz85yYeSvG++KVlUM87frUlWuvsZSa5N8v75pmQRzTh7qaotSS5Jcst8E7KoZpm9qjo5yduTPLe7n57kTXMPykKa8bXvnUm+3N2nJzk3yUfnm5IFdk2Sl6xy/qVJTp7+uSjJx+aQaU0p9gzpjCS7u/uO7v5Xki8mOXufNWcn+cz08bVJzqyqmmNGFtdB56+7v9vdD0wPb05ywpwzsphmee1Lkvdm8sOkf8wzHAttltl7fZIru/u+JOnue+eckcU1y/x1kqOmj49O8ts55mOBdff3kvxplSVnJ/lsT9yc5JiqOn4+6daGYs+QHp/k13sd3zN97oBruvvfSe5Pcuxc0rHoZpm/vV2Y5FvrmohlcdDZq6rTk5zY3d+YZzAW3iyve09J8pSquqmqbq6q1a5wwSMxy/y9J8n5VXVPkm8meeN8osEj/r5ww9k8dACW2oGuvO/7+xdnWQOHYubZqqrzk6wkef66JmJZrDp7VfWoTD569Np5BWJpzPK6tzmTragvyGSX0ver6pTu/vM6Z2PxzTJ/5yW5prs/WFXPTvK56fw9vP7xWHKj7xyu2DOke5KcuNfxCdl/y9X/1lTV5ky2Za22jQZmNcv8papelOQdSc7q7n/OKRuL7WCztyXJKUlurKpfJXlWkh1uoMcamPV99+vd/WB335lkVyZFHw7XLPN3YZIvJ0l3/yjJY5NsnUs6lt1M3xduZIo9Q/pJkpOr6qSqekwmN0nZsc+aHUkumD4+J8l3untUPz1jwzro/E23Q38ik1Lvc6aslVVnr7vv7+6t3b2tu7dlcn+Hs7p75zBxWSCzvO9+LckLk6SqtmayNf+OuaZkUc0yf3cnOTNJquppmRT7P8w1JctqR5LXTO+O/6wk93f374YO9UjYis9guvvfVfWGJNcn2ZTkU919e1VdlmRnd+9I8slMtmHtzuRK/bnDJWaRzDh/H0hyZJKvTO/ZeHd3nzVYaBbCjLMHa27G2bs+yYur6pdJHkrytu7eM1xqFsWM8/fWJFdX1Zsz2Qb9Whd0WAtV9YVMPmK0dXoPh3cneXSSdPfHM7mnw8uS7E7yQJLXDZP00JX/KwAAADBetuIDAADAiCn2AAAAMGKKPQAAAIyYYg8AAAAjptgDAADAiCn2AAAAMGKKPQAAAIzYfwCn/xUSU81vvAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 1080x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "gbm prediction progress: |████████████████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "      <th>4</th>\n",
              "      <th>mean</th>\n",
              "      <th>std</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.180</th>\n",
              "      <td>0.438444</td>\n",
              "      <td>0.413855</td>\n",
              "      <td>0.409393</td>\n",
              "      <td>0.425139</td>\n",
              "      <td>0.430297</td>\n",
              "      <td>0.423426</td>\n",
              "      <td>0.010623</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.175</th>\n",
              "      <td>0.438156</td>\n",
              "      <td>0.413638</td>\n",
              "      <td>0.408074</td>\n",
              "      <td>0.425610</td>\n",
              "      <td>0.431210</td>\n",
              "      <td>0.423338</td>\n",
              "      <td>0.011080</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.215</th>\n",
              "      <td>0.435469</td>\n",
              "      <td>0.413382</td>\n",
              "      <td>0.413919</td>\n",
              "      <td>0.426956</td>\n",
              "      <td>0.426527</td>\n",
              "      <td>0.423251</td>\n",
              "      <td>0.008464</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.210</th>\n",
              "      <td>0.435589</td>\n",
              "      <td>0.412437</td>\n",
              "      <td>0.414423</td>\n",
              "      <td>0.427968</td>\n",
              "      <td>0.425653</td>\n",
              "      <td>0.423214</td>\n",
              "      <td>0.008662</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.225</th>\n",
              "      <td>0.435107</td>\n",
              "      <td>0.412903</td>\n",
              "      <td>0.415704</td>\n",
              "      <td>0.426540</td>\n",
              "      <td>0.425695</td>\n",
              "      <td>0.423190</td>\n",
              "      <td>0.008017</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.020</th>\n",
              "      <td>0.206025</td>\n",
              "      <td>0.280270</td>\n",
              "      <td>0.224913</td>\n",
              "      <td>0.222151</td>\n",
              "      <td>0.190328</td>\n",
              "      <td>0.224737</td>\n",
              "      <td>0.030413</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.015</th>\n",
              "      <td>0.187711</td>\n",
              "      <td>0.251840</td>\n",
              "      <td>0.205190</td>\n",
              "      <td>0.198044</td>\n",
              "      <td>0.174695</td>\n",
              "      <td>0.203496</td>\n",
              "      <td>0.026262</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.010</th>\n",
              "      <td>0.168404</td>\n",
              "      <td>0.214540</td>\n",
              "      <td>0.185296</td>\n",
              "      <td>0.171950</td>\n",
              "      <td>0.158859</td>\n",
              "      <td>0.179810</td>\n",
              "      <td>0.019324</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.005</th>\n",
              "      <td>0.156899</td>\n",
              "      <td>0.167840</td>\n",
              "      <td>0.161650</td>\n",
              "      <td>0.158586</td>\n",
              "      <td>0.151825</td>\n",
              "      <td>0.159360</td>\n",
              "      <td>0.005303</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.000</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>90 rows × 7 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "              0         1         2         3         4      mean       std\n",
              "0.180  0.438444  0.413855  0.409393  0.425139  0.430297  0.423426  0.010623\n",
              "0.175  0.438156  0.413638  0.408074  0.425610  0.431210  0.423338  0.011080\n",
              "0.215  0.435469  0.413382  0.413919  0.426956  0.426527  0.423251  0.008464\n",
              "0.210  0.435589  0.412437  0.414423  0.427968  0.425653  0.423214  0.008662\n",
              "0.225  0.435107  0.412903  0.415704  0.426540  0.425695  0.423190  0.008017\n",
              "...         ...       ...       ...       ...       ...       ...       ...\n",
              "0.020  0.206025  0.280270  0.224913  0.222151  0.190328  0.224737  0.030413\n",
              "0.015  0.187711  0.251840  0.205190  0.198044  0.174695  0.203496  0.026262\n",
              "0.010  0.168404  0.214540  0.185296  0.171950  0.158859  0.179810  0.019324\n",
              "0.005  0.156899  0.167840  0.161650  0.158586  0.151825  0.159360  0.005303\n",
              "0.000  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "\n",
              "[90 rows x 7 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 58
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oT0EOn9hHE-d",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "best_f1_GBM=tune_df.sort_values(by='mean',ascending=False).index[0]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kXQ4nm44HE-f",
        "colab_type": "text"
      },
      "source": [
        "# Ensemble Tuning"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TFjJvvo9HE-g",
        "colab_type": "text"
      },
      "source": [
        "Finding the best threshold so we got the highest F1-Score using ensemble model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fMGSuU6SHE-i",
        "colab_type": "code",
        "outputId": "e0712fe8-0bb0-44a3-c069-2b7f8e67dda6",
        "colab": {}
      },
      "source": [
        "hasil_ensemble = cross_val_ensemble(df,drop_columns=[x for x in drop_columns if (x!='account_id') & (x!='is_cross_sell') & (x!='order_id') & (x!='cnt_trx') & (x!='airlines_name')],\n",
        "                                    model1='DRF',model2='GBM',f1_threshold=np.arange(0,0.45,0.005),\n",
        "                                    nfolds=5,is_feat_eng=True,is_binning=False,is_WoE=True,adjusted_WoE=True,\n",
        "                                    is_mean_encode=True,adjusted_mean=False,is_segment=False)\n",
        "hasil_ensemble.sort_values(by='mean',ascending=False)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "stackedensemble Model Build progress: |███████████████████████████████████| 100%\n",
            "stackedensemble prediction progress: |████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "stackedensemble Model Build progress: |███████████████████████████████████| 100%\n",
            "stackedensemble prediction progress: |████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "stackedensemble Model Build progress: |███████████████████████████████████| 100%\n",
            "stackedensemble prediction progress: |████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "stackedensemble Model Build progress: |███████████████████████████████████| 100%\n",
            "stackedensemble prediction progress: |████████████████████████████████████| 100%\n",
            "Done Feature Engineering!\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
            "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "gbm Model Build progress: |███████████████████████████████████████████████| 100%\n",
            "stackedensemble Model Build progress: |███████████████████████████████████| 100%\n",
            "stackedensemble prediction progress: |████████████████████████████████████| 100%\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>0</th>\n",
              "      <th>1</th>\n",
              "      <th>2</th>\n",
              "      <th>3</th>\n",
              "      <th>4</th>\n",
              "      <th>mean</th>\n",
              "      <th>std</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0.075</th>\n",
              "      <td>0.439742</td>\n",
              "      <td>0.415043</td>\n",
              "      <td>0.422600</td>\n",
              "      <td>0.427976</td>\n",
              "      <td>0.429581</td>\n",
              "      <td>0.426988</td>\n",
              "      <td>0.008152</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.085</th>\n",
              "      <td>0.438547</td>\n",
              "      <td>0.417127</td>\n",
              "      <td>0.422969</td>\n",
              "      <td>0.425672</td>\n",
              "      <td>0.429259</td>\n",
              "      <td>0.426715</td>\n",
              "      <td>0.007121</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.070</th>\n",
              "      <td>0.438356</td>\n",
              "      <td>0.414382</td>\n",
              "      <td>0.423059</td>\n",
              "      <td>0.427390</td>\n",
              "      <td>0.429505</td>\n",
              "      <td>0.426538</td>\n",
              "      <td>0.007862</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.080</th>\n",
              "      <td>0.438686</td>\n",
              "      <td>0.413699</td>\n",
              "      <td>0.422222</td>\n",
              "      <td>0.426567</td>\n",
              "      <td>0.430005</td>\n",
              "      <td>0.426236</td>\n",
              "      <td>0.008278</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.065</th>\n",
              "      <td>0.436594</td>\n",
              "      <td>0.413179</td>\n",
              "      <td>0.421477</td>\n",
              "      <td>0.426111</td>\n",
              "      <td>0.432932</td>\n",
              "      <td>0.426059</td>\n",
              "      <td>0.008308</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.025</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.020</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.015</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.010</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>0.000</th>\n",
              "      <td>0.110910</td>\n",
              "      <td>0.113242</td>\n",
              "      <td>0.112392</td>\n",
              "      <td>0.110681</td>\n",
              "      <td>0.108732</td>\n",
              "      <td>0.111191</td>\n",
              "      <td>0.001551</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>90 rows × 7 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "              0         1         2         3         4      mean       std\n",
              "0.075  0.439742  0.415043  0.422600  0.427976  0.429581  0.426988  0.008152\n",
              "0.085  0.438547  0.417127  0.422969  0.425672  0.429259  0.426715  0.007121\n",
              "0.070  0.438356  0.414382  0.423059  0.427390  0.429505  0.426538  0.007862\n",
              "0.080  0.438686  0.413699  0.422222  0.426567  0.430005  0.426236  0.008278\n",
              "0.065  0.436594  0.413179  0.421477  0.426111  0.432932  0.426059  0.008308\n",
              "...         ...       ...       ...       ...       ...       ...       ...\n",
              "0.025  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "0.020  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "0.015  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "0.010  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "0.000  0.110910  0.113242  0.112392  0.110681  0.108732  0.111191  0.001551\n",
              "\n",
              "[90 rows x 7 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 60
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_vYojYAIHE-l",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "best_f1_ensemble=hasil_ensemble.sort_values(by='mean',ascending=False).index[0]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Oc-IHaYWHE-u",
        "colab_type": "text"
      },
      "source": [
        "It can be seen that ensemble model give a higher value of F1-score compared to single model"
      ]
    }
  ]
}