preprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PREPROCESSING"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib \n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from pandas import Series, DataFrame\n",
    "import seaborn as sns\n",
    "from scipy.stats import norm\n",
    "import random as rand\n",
    "import pickle\n",
    "import re\n",
    "from itertools import groupby\n",
    "from operator import itemgetter\n",
    "\n",
    "# Environment settings\n",
    "data_path = 'Data/'\n",
    "    \n",
    "# Deserialize previously saved data from \"data-visualization\"\n",
    "dataframes = {}\n",
    "for df_name in ['test', 'train']:\n",
    "    with open(data_path+df_name+'_dv.obj', 'rb') as file:\n",
    "        dataframes[df_name] = pickle.load(file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Remove useless columns\n",
    "for df_name, df in dataframes.items():\n",
    "    dataframes[df_name] = df.drop(['Ticket','Cabin'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Transform category data into numerical\n",
    "# hack for seaborn heatmap (no support for categorical values)\n",
    "for df_name, df in dataframes.items():\n",
    "    df['Sex'] = [0 if x == 'male' else 1 for x in df['Sex']]\n",
    "    \n",
    "# One hot encoding for Embark\n",
    "for df_name, df in dataframes.items():\n",
    "    f_df = pd.get_dummies(dataframes[df_name][['Embarked']], drop_first=True)\n",
    "    dataframes[df_name].drop(['Embarked'], axis=1, inplace=True)\n",
    "    dataframes[df_name] = pd.concat([dataframes[df_name], f_df], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing 'Age' points in training set 0\n",
      "Number of missing 'Age' points in test set 0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Replace Age missing values with data generated from gaussian\n",
    "np.random.seed(666)\n",
    "for df_name, df in dataframes.items():\n",
    "    # create gaussian with available da\n",
    "    nan_filter = df['Age'].isna()\n",
    "    mu, sigma = norm.fit(df[-nan_filter]['Age'])\n",
    "    norm_dist = np.random.normal(mu, sigma, nan_filter.sum())\n",
    "    df.loc[nan_filter,'Age'] = norm_dist\n",
    "    \n",
    "    # set to zero all ages below zero\n",
    "    df.loc[df.Age<0, 'Age'] = 0\n",
    "\n",
    "# Plot histogram of the gaussian distribution of the age\n",
    "plt.hist(dataframes['train']['Age'], bins='auto')\n",
    "plt.hist(dataframes['test']['Age'], bins='auto')\n",
    "print(\"Number of missing 'Age' points in training set {}\".format(dataframes['train']['Age'].isna().sum()))\n",
    "print(\"Number of missing 'Age' points in test set {}\".format(dataframes['test']['Age'].isna().sum()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "for df_name, df in dataframes.items():\n",
    "    df['Age'] = pd.qcut(df['Age'], 6, labels=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/flaprimo/Development/Workspaces/Python/titanic-classification/venv/lib/python3.6/site-packages/matplotlib/axes/_axes.py:6462: UserWarning: The 'normed' kwarg is deprecated, and has been replaced by the 'density' kwarg.\n",
      "  warnings.warn(\"The 'normed' kwarg is deprecated, and has been \"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "152    3\n",
      "Name: Pclass, dtype: category\n",
      "Categories (3, object): [1, 2, 3]\n",
      "Number of missing 'Fare' points is in test set 0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Replace Fare missing values in test set with data generated from gaussian\n",
    "sns.distplot(dataframes['test']['Fare'][~np.isnan(dataframes['test']['Fare'])]);\n",
    "\n",
    "# Group fare by class\n",
    "grouped = dataframes['test'].groupby('Pclass').mean()\n",
    "grouped_fare = grouped['Fare']\n",
    "\n",
    "# Get row (1 row in this case) that has missing value\n",
    "nan_filter = dataframes['test']['Fare'].isna()\n",
    "print(dataframes['test'].loc[nan_filter,'Pclass'])\n",
    "\n",
    "# Since the missing value comes from a third class passenger, replace missing value\n",
    "# with mean of third class fares\n",
    "dataframes['test'].loc[nan_filter,'Fare'] = grouped_fare[2]\n",
    "print(\"Number of missing 'Fare' points is in test set {}\".format(dataframes['test']['Fare'].isna().sum()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize Fares\n",
    "max_fare = max(dataframes['train']['Fare'].max(), dataframes['test']['Fare'].max())\n",
    "for df_name, df in dataframes.items():\n",
    "    dataframes[df_name]['Fare'] = df['Fare'].apply(lambda fare: fare/max_fare)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a new feature Title which contains titles from Names\n",
    "for df_name, df in dataframes.items():\n",
    "    df['Title'] = df['Name'].apply(lambda name: re.search(' ([A-Za-z]+)\\.', name).group(1))\n",
    "    dataframes[df_name] = df.drop(['Name'], axis=1) # drop Name which is not useful anymore\n",
    "\n",
    "fig, (axis1) = plt.subplots(1,figsize=(18,6))\n",
    "sns.barplot(x='Title', y='Survived', data=dataframes['train'], ax=axis1)\n",
    "\n",
    "# get titles with corresponding survival rate\n",
    "title_survival = dataframes['train'][['Title', 'Survived']].groupby(['Title'], as_index=False) \\\n",
    ".mean().sort_values(by='Survived', ascending=False).values\n",
    "\n",
    "# round values and change tuple key\n",
    "title_survival = list(map(lambda x: (round(x[1], 1), x[0]), title_survival))\n",
    "# hack for a single woman with title 'Dona' which is only present in the test set\n",
    "# it is marked as surviving since women have larger chance to survive wrt to men\n",
    "title_survival.insert(0, (1.0, 'Dona'))\n",
    "\n",
    "# group titles by occurrency rate\n",
    "title_survival_grouped = \\\n",
    "[(k, list(list(zip(*g))[1])) for k, g in groupby(title_survival, itemgetter(0))]\n",
    "\n",
    "# get to categorical class mapper\n",
    "i = 0\n",
    "title_survival_categorical = list()\n",
    "for x in title_survival_grouped:\n",
    "    title_survival_categorical.append((i, x[1]))\n",
    "    i += 1\n",
    "\n",
    "# convert Title in train and test dataframes to categorical\n",
    "for df_name, df in dataframes.items():\n",
    "    for key, titles in title_survival_categorical:\n",
    "        df['Title'] = df['Title'].replace(titles, key)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Serialize dataframes for later use in implementations\n",
    "for df_name, df in dataframes.items():\n",
    "    with open(data_path+df_name+'_pp.obj', 'wb') as file:\n",
    "        pickle.dump(df, file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}