Chess_AdvDataCapstone.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CHESS COURSERA ADVANCED CAPSTONE PROJECT\n",
    "## How to predict who is going to win a chess match\n",
    "\n",
    "\n",
    "Guglielmo Sanchini\n",
    "\n",
    "\n",
    "The goal of this project is to predict who is going to win a chess match, in order to provide an online chess platform with useful information to use to make the platform more entertaing and matches more enjoyable\n",
    "\n",
    "Data Source: https://www.kaggle.com/datasnaek/chess"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![title](https://www.betootaadvocate.com/wp-content/uploads/2018/09/queen.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# Loading"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Importing all necessary packages and setting up Spark environment**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:36:00.387702Z",
     "start_time": "2019-10-30T08:35:57.330055Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from collections import Counter\n",
    "import joypy\n",
    "import scipy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:36:00.776747Z",
     "start_time": "2019-10-30T08:36:00.391045Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "from pyspark.ml.classification import DecisionTreeClassifier, RandomForestClassifier, LogisticRegression\n",
    "from pyspark.ml.evaluation import MulticlassClassificationEvaluator\n",
    "from sklearn.preprocessing import LabelEncoder, StandardScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from pyspark.ml.feature import OneHotEncoderEstimator, StringIndexer, StandardScaler, VectorAssembler\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.tuning import ParamGridBuilder, CrossValidator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:36:07.866439Z",
     "start_time": "2019-10-30T08:36:07.862087Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation\n",
    "from keras import optimizers, regularizers\n",
    "from keras.optimizers import Adam"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:36:09.405638Z",
     "start_time": "2019-10-30T08:36:09.392757Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "import findspark\n",
    "import os\n",
    "\n",
    "spark_location='/usr/local/Cellar/apache-spark/2.4.3/libexec/' # Set your own\n",
    "java8_location= '/Library/Java/JavaVirtualMachines/adoptopenjdk-8.jdk/Contents/Home' # Set your own\n",
    "os.environ['JAVA_HOME'] = java8_location\n",
    "findspark.init(spark_home=spark_location) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T10:49:37.133570Z",
     "start_time": "2019-10-30T10:49:37.125992Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Spark stopped\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    spark.stop()\n",
    "    print(\"Spark stopped\")\n",
    "except:\n",
    "    pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:36:34.891584Z",
     "start_time": "2019-10-30T08:36:13.141469Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from pyspark.sql import Row,SQLContext, SparkSession\n",
    "import pyspark.sql.functions as f\n",
    "from pyspark.sql.functions import *\n",
    "from pyspark.sql.types import *\n",
    "\n",
    "import pyspark\n",
    "import random\n",
    "\n",
    "sc = pyspark.SparkContext(appName=\"SPARK-CHESS\")\n",
    "spark = SparkSession.builder.appName(\"CHESS\").getOrCreate()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Loading the dataset**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:02.057702Z",
     "start_time": "2019-10-30T08:36:56.769439Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = spark.read.csv('games.csv', header=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:03.445300Z",
     "start_time": "2019-10-30T08:37:03.425223Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "        <div>\n",
       "            <p><b>SparkContext</b></p>\n",
       "\n",
       "            <p><a href=\"http://10.1.52.19:4040\">Spark UI</a></p>\n",
       "\n",
       "            <dl>\n",
       "              <dt>Version</dt>\n",
       "                <dd><code>v2.4.3</code></dd>\n",
       "              <dt>Master</dt>\n",
       "                <dd><code>local[*]</code></dd>\n",
       "              <dt>AppName</dt>\n",
       "                <dd><code>SPARK-CHESS</code></dd>\n",
       "            </dl>\n",
       "        </div>\n",
       "        "
      ],
      "text/plain": [
       "<SparkContext master=local[*] appName=SPARK-CHESS>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ETL and Feature Creation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "## Data Exploration and Cleansing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**How many rows are in the dataset?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:25.805115Z",
     "start_time": "2019-10-30T08:37:25.149239Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "20058"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**How many duplicate rows are in the dataset?**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:32.257674Z",
     "start_time": "2019-10-30T08:37:29.910673Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "19629"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = df.dropDuplicates()\n",
    "df.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Looking for NaN values**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:33.998862Z",
     "start_time": "2019-10-30T08:37:32.264096Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "19629"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.na.drop().count() #so there are no NaN values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Column names**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:35.006510Z",
     "start_time": "2019-10-30T08:37:35.000350Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['id',\n",
       " 'rated',\n",
       " 'created_at',\n",
       " 'last_move_at',\n",
       " 'turns',\n",
       " 'victory_status',\n",
       " 'winner',\n",
       " 'increment_code',\n",
       " 'white_id',\n",
       " 'white_rating',\n",
       " 'black_id',\n",
       " 'black_rating',\n",
       " 'moves',\n",
       " 'opening_eco',\n",
       " 'opening_name',\n",
       " 'opening_ply']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Explanations for some columns**\n",
    "\n",
    "increment code: https://chess.stackexchange.com/questions/18069/what-is-the-increment-in-chess\n",
    "\n",
    "how opening_name and opening_eco are related: https://www.365chess.com/eco.php"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Column types**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:38.036117Z",
     "start_time": "2019-10-30T08:37:38.029522Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('id', 'string'),\n",
       " ('rated', 'string'),\n",
       " ('created_at', 'string'),\n",
       " ('last_move_at', 'string'),\n",
       " ('turns', 'string'),\n",
       " ('victory_status', 'string'),\n",
       " ('winner', 'string'),\n",
       " ('increment_code', 'string'),\n",
       " ('white_id', 'string'),\n",
       " ('white_rating', 'string'),\n",
       " ('black_id', 'string'),\n",
       " ('black_rating', 'string'),\n",
       " ('moves', 'string'),\n",
       " ('opening_eco', 'string'),\n",
       " ('opening_name', 'string'),\n",
       " ('opening_ply', 'string')]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's drop some columns which are not relevant for our analysis**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:39.798942Z",
     "start_time": "2019-10-30T08:37:39.775868Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = df.drop(\"created_at\", \"last_move_at\", \"white_id\", \"black_id\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**In the following lines we separate the increment code column to display in two different columns the fixed time and the additional seconds given at each turn**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:41.845666Z",
     "start_time": "2019-10-30T08:37:41.755138Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#defining custom functions to split the increment_code column\n",
    "def funz1(value):\n",
    "    return value.split(\"+\")[0]\n",
    "\n",
    "def funz2(value):\n",
    "    return value.split(\"+\")[1]\n",
    "    \n",
    "udf_funz1 = udf(funz1, StringType())\n",
    "udf_funz2 = udf(funz2, StringType())\n",
    "\n",
    "#creating the two new column\n",
    "df = df.withColumn(\"increment1\", udf_funz1(\"increment_code\"))\n",
    "df = df.withColumn(\"increment2\", udf_funz2(\"increment_code\"))\n",
    "\n",
    "#dropping the old columns\n",
    "df = df.drop(\"increment_code\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We convert all wrong types into the right types, mainly from string to int**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:44.060726Z",
     "start_time": "2019-10-30T08:37:43.796635Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = df.withColumn(\"turns\",df[\"turns\"].cast(\"int\"))\n",
    "df = df.withColumn(\"white_rating\",df[\"white_rating\"].cast(\"int\"))\n",
    "df = df.withColumn(\"black_rating\",df[\"black_rating\"].cast(\"int\"))\n",
    "df = df.withColumn(\"opening_ply\",df[\"opening_ply\"].cast(\"int\"))\n",
    "df = df.withColumn(\"increment1\",df[\"increment1\"].cast(\"int\"))\n",
    "df = df.withColumn(\"increment2\",df[\"increment2\"].cast(\"int\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We inspect the new types, to check if they are right**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:45.983607Z",
     "start_time": "2019-10-30T08:37:45.974160Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('id', 'string'),\n",
       " ('rated', 'string'),\n",
       " ('turns', 'int'),\n",
       " ('victory_status', 'string'),\n",
       " ('winner', 'string'),\n",
       " ('white_rating', 'int'),\n",
       " ('black_rating', 'int'),\n",
       " ('moves', 'string'),\n",
       " ('opening_eco', 'string'),\n",
       " ('opening_name', 'string'),\n",
       " ('opening_ply', 'int'),\n",
       " ('increment1', 'int'),\n",
       " ('increment2', 'int')]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's examine the rated column**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:51.009634Z",
     "start_time": "2019-10-30T08:37:47.998231Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+\n",
      "|rated|count|\n",
      "+-----+-----+\n",
      "|FALSE| 1852|\n",
      "|False| 1960|\n",
      "| TRUE| 7418|\n",
      "| True| 8399|\n",
      "+-----+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupBy('rated').count().show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**This column needs cleaning, which is performed in the cell below**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:54.423857Z",
     "start_time": "2019-10-30T08:37:54.330096Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = df.withColumn(\"rated\", \\\n",
    "              when(df[\"rated\"] == \"FALSE\", \"False\").otherwise(df[\"rated\"]))\n",
    "\n",
    "df = df.withColumn(\"rated\", \\\n",
    "              when(df[\"rated\"] == \"TRUE\", \"True\").otherwise(df[\"rated\"]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's review the updated column**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:37:57.459462Z",
     "start_time": "2019-10-30T08:37:55.249166Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+\n",
      "|rated|count|\n",
      "+-----+-----+\n",
      "|False| 3812|\n",
      "| True|15817|\n",
      "+-----+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupBy('rated').count().show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's redrop duplicates after those transformations, to catch other possible identical rows**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:38:12.063543Z",
     "start_time": "2019-10-30T08:38:01.118528Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19629\n",
      "19113\n"
     ]
    }
   ],
   "source": [
    "print(df.count())\n",
    "df = df.dropDuplicates()\n",
    "print(df.count())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's visualize the rated column**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:16:42.882595Z",
     "start_time": "2019-10-28T11:16:35.852564Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.groupBy('rated').count().toPandas().plot.pie(y=\"count\", figsize=(5, 5), \n",
    "                                                labels=[\"False\", \"True\"], autopct='%.1f', \n",
    "                                                colors=[\"r\", \"g\"], pctdistance=0.5,\n",
    "                                               textprops={\"size\":16}, explode=[0.1,0.1],\n",
    "                                               shadow = True)\n",
    "plt.ylabel(\"\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's visualize the victory status**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:16:49.698364Z",
     "start_time": "2019-10-28T11:16:42.886302Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------+-----+\n",
      "|victory_status|count|\n",
      "+--------------+-----+\n",
      "|        resign|10695|\n",
      "|     outoftime| 1598|\n",
      "|          mate| 5974|\n",
      "|          draw|  846|\n",
      "+--------------+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupBy('victory_status').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:16:56.576622Z",
     "start_time": "2019-10-28T11:16:49.706877Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.groupBy('victory_status').count().toPandas().plot.pie(y=\"count\", figsize=(6, 6), \n",
    "                                                labels=[\"resign\", \"outoftime\", \"mate\", \"draw\"], autopct='%.1f', \n",
    "                                                colors=[\"gold\", \"orange\", \"lightblue\", \"lime\"], pctdistance=0.5,\n",
    "                                               textprops={\"size\":16}, legend=None, explode=[0.05,0.05,0.05,0.05],\n",
    "                                               shadow = True)\n",
    "plt.ylabel(\"\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see how winners are distributed: we see that the white player (who moves first) has a slight advantage over the black one;**\n",
    "\n",
    "**this is a well-known fact in the chess world**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:17:03.376307Z",
     "start_time": "2019-10-28T11:16:56.580977Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----+\n",
      "|winner|count|\n",
      "+------+-----+\n",
      "| white| 9545|\n",
      "| black| 8680|\n",
      "|  draw|  888|\n",
      "+------+-----+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupBy('winner').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:17:10.276940Z",
     "start_time": "2019-10-28T11:17:03.379572Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.groupBy('winner').count().toPandas().plot.pie(y=\"count\", figsize=(5, 5), \n",
    "                                                labels=[\"White\", \"Black\", \"Draw\"], autopct='%.1f', \n",
    "                                                colors=[\"ivory\", \"slategrey\", \"springgreen\"], pctdistance=0.5,\n",
    "                                               textprops={\"size\":16}, explode=[0.1,0.1, 0.1],\n",
    "                                               shadow = True)\n",
    "plt.ylabel(\"\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We now inspect a bit the column of the moves**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:17:15.241968Z",
     "start_time": "2019-10-28T11:17:10.281014Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+\n",
      "|               moves|\n",
      "+--------------------+\n",
      "|d4 c5 c3 cxd4 cxd...|\n",
      "|e4 c5 c3 Nc6 d4 c...|\n",
      "|e4 e5 Nf3 Nc6 d4 ...|\n",
      "|d4 d5 Nf3 Nf6 Bf4...|\n",
      "|e4 Nc6 d4 Nf6 e5 ...|\n",
      "|d4 Nf6 Nc3 d6 e4 ...|\n",
      "|d4 e6 c4 c5 d5 Qb...|\n",
      "|d4 f5 e3 Nf6 Nf3 ...|\n",
      "|d4 d5 c4 c6 Nf3 N...|\n",
      "|e4 e5 Nf3 Nc6 Bb5...|\n",
      "|d4 b6 Nf3 Bb7 e3 ...|\n",
      "|e4 e5 Nf3 d6 Nc3 ...|\n",
      "|e4 c5 Nf3 d6 d4 c...|\n",
      "|e4 e5 Nf3 Nc6 Bc4...|\n",
      "|e4 e5 Nf3 Nc6 Bc4...|\n",
      "|e4 Nc6 Bc4 Nf6 Nf...|\n",
      "|d3 e6 e3 Nc6 Nf3 ...|\n",
      "|e4 Nc6 d4 d5 e5 B...|\n",
      "|e4 h5 Nf3 g6 d4 e...|\n",
      "|e4 c5 Nf3 d6 Nc3 ...|\n",
      "+--------------------+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\"moves\").show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T11:17:19.954703Z",
     "start_time": "2019-10-28T11:17:15.245607Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+--------------------+---+----+\n",
      "|      id|               moves|pos| val|\n",
      "+--------+--------------------+---+----+\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  0|  d4|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  1|  c5|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  2|  c3|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  3|cxd4|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  4|cxd4|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  5| Nf6|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  6| Nc3|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  7|  g6|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  8|  b3|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...|  9| Bg7|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 10| Bb2|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 11| O-O|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 12|  e3|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 13|  e6|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 14| Nf3|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 15|  a6|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 16| Be2|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 17|  d5|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 18| O-O|\n",
      "|YopBjBqM|[d4, c5, c3, cxd4...| 19| Nc6|\n",
      "+--------+--------------------+---+----+\n",
      "only showing top 20 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\n",
    "        \"id\",\n",
    "        f.split(\"moves\", \" \").alias(\"moves\"),\n",
    "        f.posexplode(f.split(\"moves\", \" \")).alias(\"pos\", \"val\")\n",
    "        ).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**To be able to better analyze our dataset, we first create a new column storing the first two moves and the last two moves, and then split them so that we end up with 4 columns, each for every move that we kept**\n",
    "\n",
    "After some experiments keeping all the moves or just subsets, I ended up choosing the first two and last two as this configuration lead to the best results afterwards, and it was also a good compromise concerning the execution speed of neural nets and the other algorithms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:02.249852Z",
     "start_time": "2019-10-30T08:41:02.211432Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#function that retrieves the first two and last two moves from the column moves\n",
    "def funz_moves(value):\n",
    "    split_moves = value.split(\" \")\n",
    "    \n",
    "    if len(split_moves) >= 4:\n",
    "        fin = split_moves[:2] + split_moves[-2:]\n",
    "        fin = \" \".join(fin)\n",
    "        return fin\n",
    "    else:\n",
    "        split_moves = \" \".join(split_moves)\n",
    "        return split_moves\n",
    "\n",
    "udf_funz_moves = udf(funz_moves, StringType())\n",
    "\n",
    "#creating the new column\n",
    "df = df.withColumn(\"moves_new\", udf_funz_moves(\"moves\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:11.346098Z",
     "start_time": "2019-10-30T08:41:05.254786Z"
    },
    "hidden": true,
    "scrolled": false
   },
   "outputs": [
    {
     "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>id</th>\n",
       "      <th>rated</th>\n",
       "      <th>turns</th>\n",
       "      <th>victory_status</th>\n",
       "      <th>winner</th>\n",
       "      <th>white_rating</th>\n",
       "      <th>black_rating</th>\n",
       "      <th>moves</th>\n",
       "      <th>opening_eco</th>\n",
       "      <th>opening_name</th>\n",
       "      <th>opening_ply</th>\n",
       "      <th>increment1</th>\n",
       "      <th>increment2</th>\n",
       "      <th>moves_new</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>YopBjBqM</td>\n",
       "      <td>True</td>\n",
       "      <td>139</td>\n",
       "      <td>draw</td>\n",
       "      <td>draw</td>\n",
       "      <td>1736</td>\n",
       "      <td>1893</td>\n",
       "      <td>d4 c5 c3 cxd4 cxd4 Nf6 Nc3 g6 b3 Bg7 Bb2 O-O e...</td>\n",
       "      <td>A43</td>\n",
       "      <td>Old Benoni Defense</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>d4 c5 Rh3 Kd2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>SipCSO1G</td>\n",
       "      <td>True</td>\n",
       "      <td>96</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>2209</td>\n",
       "      <td>2044</td>\n",
       "      <td>e4 c5 c3 Nc6 d4 cxd4 cxd4 e6 Nc3 Bb4 Ne2 Nge7 ...</td>\n",
       "      <td>B22</td>\n",
       "      <td>Sicilian Defense: Alapin Variation</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>e4 c5 Re2 a3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tpi9ti3J</td>\n",
       "      <td>True</td>\n",
       "      <td>58</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>1652</td>\n",
       "      <td>1624</td>\n",
       "      <td>e4 e5 Nf3 Nc6 d4 exd4 Nxd4 Bb4+ c3 Ba5 Nf5 Qf6...</td>\n",
       "      <td>C45</td>\n",
       "      <td>Scotch Game: Malaniuk Variation</td>\n",
       "      <td>8</td>\n",
       "      <td>6</td>\n",
       "      <td>8</td>\n",
       "      <td>e4 e5 h3 Bd5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         id rated  turns victory_status winner  white_rating  black_rating  \\\n",
       "0  YopBjBqM  True    139           draw   draw          1736          1893   \n",
       "1  SipCSO1G  True     96         resign  black          2209          2044   \n",
       "2  tpi9ti3J  True     58         resign  black          1652          1624   \n",
       "\n",
       "                                               moves opening_eco  \\\n",
       "0  d4 c5 c3 cxd4 cxd4 Nf6 Nc3 g6 b3 Bg7 Bb2 O-O e...         A43   \n",
       "1  e4 c5 c3 Nc6 d4 cxd4 cxd4 e6 Nc3 Bb4 Ne2 Nge7 ...         B22   \n",
       "2  e4 e5 Nf3 Nc6 d4 exd4 Nxd4 Bb4+ c3 Ba5 Nf5 Qf6...         C45   \n",
       "\n",
       "                         opening_name  opening_ply  increment1  increment2  \\\n",
       "0                  Old Benoni Defense            2          10           0   \n",
       "1  Sicilian Defense: Alapin Variation            3           8           0   \n",
       "2     Scotch Game: Malaniuk Variation            8           6           8   \n",
       "\n",
       "       moves_new  \n",
       "0  d4 c5 Rh3 Kd2  \n",
       "1   e4 c5 Re2 a3  \n",
       "2   e4 e5 h3 Bd5  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#let's see some rows of the new dataframe\n",
    "df.limit(3).toPandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:21.534016Z",
     "start_time": "2019-10-30T08:41:11.351909Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#here we create the new four columns containing the moves we want to keep\n",
    "df_split2 = df.select(\n",
    "        \"id\",\n",
    "        f.split(\"moves_new\", \" \").alias(\"moves_new\"),\n",
    "        f.posexplode(f.split(\"moves_new\", \" \")).alias(\"pos\", \"val\")\n",
    "    )\\\n",
    "    .drop(\"val\")\\\n",
    "    .select(\n",
    "        \"id\",\n",
    "        f.concat(f.lit(\"moves_new\"),f.col(\"pos\").cast(\"string\")).alias(\"name\"),\n",
    "        f.expr(\"moves_new[pos]\").alias(\"val\")\n",
    "    )\\\n",
    "    .groupBy(\"id\").pivot(\"name\").agg(f.first(\"val\"))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:24.756150Z",
     "start_time": "2019-10-30T08:41:24.750248Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "column_list = [\"id\"] + [\"moves_new{0}\".format(i) for i in range(4)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:25.107442Z",
     "start_time": "2019-10-30T08:41:25.085020Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df_split2 = df_split2.select(column_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:25.682461Z",
     "start_time": "2019-10-30T08:41:25.605673Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#we join the old dataframe with the new one with the four columns\n",
    "df = df.join(df_split2, [\"id\"])\n",
    "df = df.drop(\"moves\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see the shape of the new dataframe**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:42.705846Z",
     "start_time": "2019-10-30T08:41:28.085798Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19113\n",
      "17\n"
     ]
    }
   ],
   "source": [
    "print(df.count())\n",
    "print(len(df.columns))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see some summary statistics and a histogram to get some insights from the TURN feature**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:47:55.974450Z",
     "start_time": "2019-10-28T12:47:44.987707Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+------------------+\n",
      "|summary|             turns|\n",
      "+-------+------------------+\n",
      "|  count|             19113|\n",
      "|   mean|60.513838748495786|\n",
      "| stddev| 33.48826396268022|\n",
      "|    min|                 1|\n",
      "|    max|               349|\n",
      "+-------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\"turns\").describe().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:14:03.932220Z",
     "start_time": "2019-10-28T10:13:52.272357Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.select(\"turns\").toPandas().plot(kind=\"hist\", figsize=(15,5), bins=40, edgecolor=\"black\", alpha=0.8, legend=None)\n",
    "plt.title(\"Distribution of the number of Turns\")\n",
    "plt.xlabel(\"Turns\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see some summary statistics and a histogram to get some insights from the OPENING_PLAY feature**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:14:15.254082Z",
     "start_time": "2019-10-28T10:14:03.940870Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+------------------+\n",
      "|summary|       opening_ply|\n",
      "+-------+------------------+\n",
      "|  count|             19113|\n",
      "|   mean| 4.815779835713912|\n",
      "| stddev|2.7982829080050102|\n",
      "|    min|                 1|\n",
      "|    max|                28|\n",
      "+-------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\"opening_ply\").describe().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:14:29.048618Z",
     "start_time": "2019-10-28T10:14:15.257817Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.select(\"opening_ply\").toPandas().plot(kind=\"hist\", figsize=(15,5), bins=27, color=\"lime\",\n",
    "                                         edgecolor=\"black\", alpha=0.4, legend=None)\n",
    "plt.title(\"Distribution of the number of moves in the opening phase\")\n",
    "plt.xticks([1.5, 2.5, 3.5, 4.5, 5.5, 6.5, 7.5, 8.5, 9.5, 10.5,11.5,12.5,13.5,14.5,15.5,20.5, 25.5,28.5], \n",
    "           [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,20,25,28])\n",
    "plt.xlabel(\"n° moves\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:14:41.019204Z",
     "start_time": "2019-10-28T10:14:29.054162Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.colors as mcolors\n",
    "clist = [(0, \"red\"), (0.125, \"red\"), (0.25, \"orange\"), (0.5, \"green\"), \n",
    "         (0.7, \"green\"), (0.75, \"blue\"), (1, \"blue\")]\n",
    "rvb = mcolors.LinearSegmentedColormap.from_list(\"\", clist)\n",
    "\n",
    "N = 50\n",
    "x = np.arange(N).astype(float)\n",
    "y = np.random.uniform(0, 5, size=(N,))\n",
    "\n",
    "\n",
    "df.select(\"opening_eco\").toPandas()[\"opening_eco\"].value_counts().iloc[:N].plot(kind=\"bar\", \n",
    "                                                                                 figsize=(16,5),\n",
    "                                                                                 color=rvb(x/N) )\n",
    "plt.title(\"Most frequent opening plays\")\n",
    "plt.xlabel(\"Opening play code\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see some summary statistics and a histogram to get some insights from the INCREMENT1 feature**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:14:52.137441Z",
     "start_time": "2019-10-28T10:14:41.026532Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+------------------+\n",
      "|summary|        increment1|\n",
      "+-------+------------------+\n",
      "|  count|             19113|\n",
      "|   mean|13.785277036571967|\n",
      "| stddev|17.072446733601073|\n",
      "|    min|                 0|\n",
      "|    max|               180|\n",
      "+-------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\"increment1\").describe().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:15:03.152908Z",
     "start_time": "2019-10-28T10:14:52.142566Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.select(\"increment1\").toPandas().plot(kind=\"hist\", figsize=(15,5), color=\"blueviolet\",\n",
    "                                        bins=25, edgecolor=\"black\", alpha=0.8, legend=None)\n",
    "plt.title(\"Distribution of increment1\")\n",
    "plt.xlabel(\"Increment1 (min)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see some summary statistics and a histogram to get some insights from the INCREMENT2 feature**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:15:13.748690Z",
     "start_time": "2019-10-28T10:15:03.159916Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-------+------------------+\n",
      "|summary|        increment2|\n",
      "+-------+------------------+\n",
      "|  count|             19113|\n",
      "|   mean|5.1464971485376445|\n",
      "| stddev|13.808620253370899|\n",
      "|    min|                 0|\n",
      "|    max|               180|\n",
      "+-------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.select(\"increment2\").describe().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:15:25.896486Z",
     "start_time": "2019-10-28T10:15:13.754993Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.select(\"increment2\").toPandas().plot(kind=\"hist\", figsize=(15,5), color=\"burlywood\",\n",
    "                                        bins=25, edgecolor=\"black\", alpha=0.8, legend=None)\n",
    "plt.title(\"Distribution of increment2\")\n",
    "plt.xlabel(\"Increment2 (sec)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's explore how the winner columns and the player ratings are related, using joyplots**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:15:38.761189Z",
     "start_time": "2019-10-28T10:15:25.917207Z"
    },
    "hidden": true,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Draw Plot\n",
    "fig, axes = joypy.joyplot(df.select(\"white_rating\", \"black_rating\", \"winner\").toPandas(), \n",
    "                          column=['white_rating', 'black_rating'],by = \"winner\", ylim='own', \n",
    "                          color=[\"grey\", \"olive\"], legend=True,\n",
    "                          figsize=(15,6), alpha=0.5)\n",
    "\n",
    "# Decoration\n",
    "\n",
    "plt.title('Joy Plot of Black and White Rating', fontsize=22)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:48:33.561558Z",
     "start_time": "2019-10-28T12:48:21.223987Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+------------------+------------------+\n",
      "|winner| avg(white_rating)| avg(black_rating)|\n",
      "+------+------------------+------------------+\n",
      "| white|1634.7327396542692|1540.4346778418019|\n",
      "| black|1550.8154377880185| 1639.265668202765|\n",
      "|  draw|1649.3164414414414|1642.1903153153153|\n",
      "+------+------------------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupby(\"winner\").mean().select([\"winner\",\"avg(white_rating)\", \"avg(black_rating)\"]).show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We clearly see from the plots and the box above that there is a correlation between the rating of the players and the winner of the match**\n",
    "\n",
    "**We formalize this intuition with the following t-test: we see that p-values are really low when there is a winner, so the difference in rating of players is statistically significant; in case of a draw, the p-value is high, so the difference of player ratings is not statistically significant**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:15:58.584181Z",
     "start_time": "2019-10-28T10:15:58.494466Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#create a temp view to perform SQL queries\n",
    "df.createTempView(\"dataframe\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:16:45.327435Z",
     "start_time": "2019-10-28T10:16:00.083151Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#subsetting player ratings based on the match outcome\n",
    "df_tt1 = spark.sql(\"select white_rating, black_rating from dataframe where winner = '{0}'\".format(\"white\")).toPandas()\n",
    "df_tt2 = spark.sql(\"select white_rating, black_rating from dataframe where winner = '{0}'\".format(\"black\")).toPandas()\n",
    "df_tt3 = spark.sql(\"select white_rating, black_rating from dataframe where winner = '{0}'\".format(\"draw\")).toPandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:16:45.343059Z",
     "start_time": "2019-10-28T10:16:45.329886Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Ttest_indResult(statistic=22.837035036935784, pvalue=6.591467723433094e-114)\n",
      "Ttest_indResult(statistic=-20.374237387648417, pvalue=3.2960912165480816e-91)\n",
      "Ttest_indResult(statistic=0.49189177002182255, pvalue=0.6228567156066165)\n"
     ]
    }
   ],
   "source": [
    "#performing t-tests to determine if the differences in the means of player ratings are statistically significant, \n",
    "#depending on the outcomes\n",
    "print(scipy.stats.ttest_ind(df_tt1.white_rating, df_tt1.black_rating))\n",
    "print(scipy.stats.ttest_ind(df_tt2.white_rating, df_tt2.black_rating))\n",
    "print(scipy.stats.ttest_ind(df_tt3.white_rating, df_tt3.black_rating))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**In the following cells we explore a bit how the columns OPENING_NAME and OPENING_ECO are related**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:05.812536Z",
     "start_time": "2019-10-28T10:17:41.384667Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "ct = pd.crosstab(df.select(\"opening_eco\").toPandas()[\"opening_eco\"], \n",
    "                 df.select(\"opening_name\").toPandas()[\"opening_name\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:05.847177Z",
     "start_time": "2019-10-28T10:18:05.815596Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>opening_name</th>\n",
       "      <th>Alekhine Defense</th>\n",
       "      <th>Alekhine Defense #2</th>\n",
       "      <th>Alekhine Defense #3</th>\n",
       "      <th>Alekhine Defense: Balogh Variation</th>\n",
       "      <th>Alekhine Defense: Brooklyn Variation</th>\n",
       "      <th>Alekhine Defense: Exchange Variation</th>\n",
       "      <th>Alekhine Defense: Four Pawns Attack</th>\n",
       "      <th>Alekhine Defense: Four Pawns Attack |  6...Nc6</th>\n",
       "      <th>Alekhine Defense: Four Pawns Attack |  Fianchetto Variation</th>\n",
       "      <th>Alekhine Defense: Four Pawns Attack |  Main Line</th>\n",
       "      <th>...</th>\n",
       "      <th>Zukertort Opening: Quiet System</th>\n",
       "      <th>Zukertort Opening: Reversed Mexican Defense</th>\n",
       "      <th>Zukertort Opening: Ross Gambit</th>\n",
       "      <th>Zukertort Opening: Santasiere's Folly</th>\n",
       "      <th>Zukertort Opening: Sicilian Invitation</th>\n",
       "      <th>Zukertort Opening: Slav Invitation</th>\n",
       "      <th>Zukertort Opening: Tennison Gambit</th>\n",
       "      <th>Zukertort Opening: The Walrus</th>\n",
       "      <th>Zukertort Opening: Vos Gambit</th>\n",
       "      <th>Zukertort Opening: Wade Defense</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>opening_eco</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>A00</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A01</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 1477 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "opening_name  Alekhine Defense  Alekhine Defense #2  Alekhine Defense #3  \\\n",
       "opening_eco                                                                \n",
       "A00                          0                    0                    0   \n",
       "A01                          0                    0                    0   \n",
       "\n",
       "opening_name  Alekhine Defense: Balogh Variation  \\\n",
       "opening_eco                                        \n",
       "A00                                            0   \n",
       "A01                                            0   \n",
       "\n",
       "opening_name  Alekhine Defense: Brooklyn Variation  \\\n",
       "opening_eco                                          \n",
       "A00                                              0   \n",
       "A01                                              0   \n",
       "\n",
       "opening_name  Alekhine Defense: Exchange Variation  \\\n",
       "opening_eco                                          \n",
       "A00                                              0   \n",
       "A01                                              0   \n",
       "\n",
       "opening_name  Alekhine Defense: Four Pawns Attack  \\\n",
       "opening_eco                                         \n",
       "A00                                             0   \n",
       "A01                                             0   \n",
       "\n",
       "opening_name  Alekhine Defense: Four Pawns Attack |  6...Nc6  \\\n",
       "opening_eco                                                    \n",
       "A00                                                        0   \n",
       "A01                                                        0   \n",
       "\n",
       "opening_name  Alekhine Defense: Four Pawns Attack |  Fianchetto Variation  \\\n",
       "opening_eco                                                                 \n",
       "A00                                                           0             \n",
       "A01                                                           0             \n",
       "\n",
       "opening_name  Alekhine Defense: Four Pawns Attack |  Main Line  ...  \\\n",
       "opening_eco                                                     ...   \n",
       "A00                                                          0  ...   \n",
       "A01                                                          0  ...   \n",
       "\n",
       "opening_name  Zukertort Opening: Quiet System  \\\n",
       "opening_eco                                     \n",
       "A00                                         0   \n",
       "A01                                         0   \n",
       "\n",
       "opening_name  Zukertort Opening: Reversed Mexican Defense  \\\n",
       "opening_eco                                                 \n",
       "A00                                                     0   \n",
       "A01                                                     0   \n",
       "\n",
       "opening_name  Zukertort Opening: Ross Gambit  \\\n",
       "opening_eco                                    \n",
       "A00                                        0   \n",
       "A01                                        0   \n",
       "\n",
       "opening_name  Zukertort Opening: Santasiere's Folly  \\\n",
       "opening_eco                                           \n",
       "A00                                               0   \n",
       "A01                                               0   \n",
       "\n",
       "opening_name  Zukertort Opening: Sicilian Invitation  \\\n",
       "opening_eco                                            \n",
       "A00                                                0   \n",
       "A01                                                0   \n",
       "\n",
       "opening_name  Zukertort Opening: Slav Invitation  \\\n",
       "opening_eco                                        \n",
       "A00                                            0   \n",
       "A01                                            0   \n",
       "\n",
       "opening_name  Zukertort Opening: Tennison Gambit  \\\n",
       "opening_eco                                        \n",
       "A00                                            0   \n",
       "A01                                            0   \n",
       "\n",
       "opening_name  Zukertort Opening: The Walrus  Zukertort Opening: Vos Gambit  \\\n",
       "opening_eco                                                                  \n",
       "A00                                       0                              0   \n",
       "A01                                       0                              0   \n",
       "\n",
       "opening_name  Zukertort Opening: Wade Defense  \n",
       "opening_eco                                    \n",
       "A00                                         0  \n",
       "A01                                         0  \n",
       "\n",
       "[2 rows x 1477 columns]"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ct.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:06.011892Z",
     "start_time": "2019-10-28T10:18:05.852958Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "list_incongr = []\n",
    "z = ct.apply(Counter).values\n",
    "for i, elem in enumerate(z):\n",
    "    if elem[0]!=364:\n",
    "        list_incongr.append(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:06.023495Z",
     "start_time": "2019-10-28T10:18:06.015424Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "ct2 = ct.iloc[:, list_incongr].T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:06.192390Z",
     "start_time": "2019-10-28T10:18:06.031987Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "list_incongr_col = [i for i in range(365) if ct2.sum().iloc[i]!=0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T10:18:06.230623Z",
     "start_time": "2019-10-28T10:18:06.196606Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "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>opening_eco</th>\n",
       "      <th>A02</th>\n",
       "      <th>A03</th>\n",
       "      <th>A04</th>\n",
       "      <th>A06</th>\n",
       "      <th>A07</th>\n",
       "      <th>A08</th>\n",
       "      <th>A09</th>\n",
       "      <th>A20</th>\n",
       "      <th>A21</th>\n",
       "      <th>A34</th>\n",
       "      <th>...</th>\n",
       "      <th>D45</th>\n",
       "      <th>D46</th>\n",
       "      <th>D52</th>\n",
       "      <th>D74</th>\n",
       "      <th>D76</th>\n",
       "      <th>E01</th>\n",
       "      <th>E06</th>\n",
       "      <th>E12</th>\n",
       "      <th>E91</th>\n",
       "      <th>E94</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>opening_name</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Alekhine Defense</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Bird Opening</th>\n",
       "      <td>58</td>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Borg Defense: Borg Gambit</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 101 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "opening_eco                A02  A03  A04  A06  A07  A08  A09  A20  A21  A34  \\\n",
       "opening_name                                                                  \n",
       "Alekhine Defense             0    0    0    0    0    0    0    0    0    0   \n",
       "Bird Opening                58   11    0    0    0    0    0    0    0    0   \n",
       "Borg Defense: Borg Gambit    0    0    0    0    0    0    0    0    0    0   \n",
       "\n",
       "opening_eco                ...  D45  D46  D52  D74  D76  E01  E06  E12  E91  \\\n",
       "opening_name               ...                                                \n",
       "Alekhine Defense           ...    0    0    0    0    0    0    0    0    0   \n",
       "Bird Opening               ...    0    0    0    0    0    0    0    0    0   \n",
       "Borg Defense: Borg Gambit  ...    0    0    0    0    0    0    0    0    0   \n",
       "\n",
       "opening_eco                E94  \n",
       "opening_name                    \n",
       "Alekhine Defense             0  \n",
       "Bird Opening                 0  \n",
       "Borg Defense: Borg Gambit    0  \n",
       "\n",
       "[3 rows x 101 columns]"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ct3 = ct.iloc[list_incongr_col, list_incongr].T\n",
    "ct3.head(3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We now draw the barplots of the first two moves of games, to see how they related with the match outcome**\n",
    "\n",
    "**We notice in the plots below that the difference in moves with respect to different match outcomes are not strong enough to consider a particular move a good indicator of the final winner**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:49:14.422311Z",
     "start_time": "2019-10-28T12:48:57.278460Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#fuction that plots the bars with colors and labels\n",
    "def subcategorybar(X, vals, width=0.8):\n",
    "    plt.figure(figsize=(15,5))\n",
    "    n = len(vals)\n",
    "    _X = np.arange(len(X))\n",
    "    labels=[\"white\", \"black\"]\n",
    "    for i in range(n):\n",
    "        plt.bar(_X - width/2. + i/float(n)*width, vals[i]/vals[i].sum(), \n",
    "                width=width/float(n), align=\"edge\", label=labels[i], \n",
    "                color=labels[i], edgecolor=\"black\")   \n",
    "    plt.xticks(_X, X)\n",
    "    plt.legend()\n",
    "\n",
    "# function that provides the count of the moves divide by match outcome and integrates them to solve\n",
    "# compatibily issues deriving from moves that are never used by a white or black opponent\n",
    "def moves_win(name_col):\n",
    "    df_pd = df.select(name_col, \"winner\").toPandas()    \n",
    "\n",
    "    X = list(df_pd[df_pd.winner == \"white\"][\"{0}\".format(name_col)].value_counts().index)\n",
    "    X2 = list(df_pd[df_pd.winner == \"black\"][\"{0}\".format(name_col)].value_counts().index)\n",
    "    \n",
    "    list_uncommon1 = [i for i in X if i not in X2] \n",
    "    list_uncommon2 = [i for i in X2 if i not in X]\n",
    "    \n",
    "    X_mod = X + list_uncommon2\n",
    "    \n",
    "    if len(list_uncommon2) == 0:\n",
    "        \n",
    "        Y = df_pd[df_pd.winner == \"white\"][\"{0}\".format(name_col)].value_counts().reindex(X_mod).values\n",
    "    \n",
    "    else:\n",
    "        \n",
    "        for i in range(len(list_uncommon2)):\n",
    "            Y = df_pd[df_pd.winner == \"white\"][\"{0}\".format(name_col)].\\\n",
    "            value_counts().append(pd.Series([0], [list_uncommon2[i]]))\n",
    "        \n",
    "        Y = Y.reindex(X_mod).values\n",
    "    \n",
    "    if len(list_uncommon1) == 0:\n",
    "        \n",
    "        Z = df_pd[df_pd.winner == \"black\"][\"{0}\".format(name_col)].value_counts().reindex(X_mod).values\n",
    "    \n",
    "    else:\n",
    "        \n",
    "        for i in range(len(list_uncommon1)):\n",
    "            Z = df_pd[df_pd.winner == \"black\"][\"{0}\".format(name_col)].\\\n",
    "            value_counts().append(pd.Series([0], [list_uncommon1[i]]))\n",
    "        \n",
    "        Z = Z.reindex(X_mod).values\n",
    "\n",
    "    subcategorybar(X_mod, [Y,Z])\n",
    "    \n",
    "    plt.show()\n",
    "    \n",
    "moves_win(\"moves_new0\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:49:34.288941Z",
     "start_time": "2019-10-28T12:49:17.674162Z"
    },
    "hidden": true,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "moves_win(\"moves_new1\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**In the following boxes we clearly see that the number of turns played is strongly related both to the outcome of a match and to the victory_status**\n",
    "\n",
    "In particular it is clear that matches ending in a draw are characterized by a high number of turns played"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:49:46.771655Z",
     "start_time": "2019-10-28T12:49:34.293147Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----------------+------------------+\n",
      "|winner| avg(opening_ply)|        avg(turns)|\n",
      "+------+-----------------+------------------+\n",
      "| white|4.849449973808277| 57.77852278679937|\n",
      "| black|4.748271889400922|60.795852534562215|\n",
      "|  draw|5.113738738738738| 87.15878378378379|\n",
      "+------+-----------------+------------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupby(\"winner\").avg().select(\"winner\", \"avg(opening_ply)\", \"avg(turns)\").show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-28T12:49:58.819438Z",
     "start_time": "2019-10-28T12:49:46.775568Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------+-----------------+-----------------+\n",
      "|victory_status| avg(opening_ply)|       avg(turns)|\n",
      "+--------------+-----------------+-----------------+\n",
      "|        resign|4.993174380551659|53.91182795698925|\n",
      "|     outoftime|4.677096370463079|72.97872340425532|\n",
      "|          mate|4.492132574489454|65.56896551724138|\n",
      "|          draw|5.120567375886525|84.73404255319149|\n",
      "+--------------+-----------------+-----------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "df.groupby(\"victory_status\").avg().select(\"victory_status\", \"avg(opening_ply)\", \"avg(turns)\").show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We drop other columns that are not needed for the analysis**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:41:57.414762Z",
     "start_time": "2019-10-30T08:41:57.389260Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "df = df.drop(\"id\", \"moves_new\", \"opening_eco\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We notice that the moves_new1, moves_new2 and moves_new3 columsn have some NaNs deriving from that fact that some games terminated after the first, second or third turn; we fill NaN values with 0 to avoid problems in the following phase of modeling; so 0 corresponds to \"NO_MOVE\", and is an acceptable and consistent value to assign to those columns**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:42:24.233728Z",
     "start_time": "2019-10-30T08:41:59.579756Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+--------------+------+------------+------------+------------+-----------+----------+----------+----------+----------+----------+----------+\n",
      "|rated|turns|victory_status|winner|white_rating|black_rating|opening_name|opening_ply|increment1|increment2|moves_new0|moves_new1|moves_new2|moves_new3|\n",
      "+-----+-----+--------------+------+------------+------------+------------+-----------+----------+----------+----------+----------+----------+----------+\n",
      "|    0|    0|             0|     0|           0|           0|           0|          0|         0|         0|         0|        17|       194|       276|\n",
      "+-----+-----+--------------+------+------------+------------+------------+-----------+----------+----------+----------+----------+----------+----------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#check for NaNs and Nulls in every column of df\n",
    "df.select([count(when(isnan(c) | col(c).isNull(), c)).alias(c) for c in df.columns]).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:42:24.318886Z",
     "start_time": "2019-10-30T08:42:24.241313Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#filling null values with 0\n",
    "df = df.fillna({'moves_new1':0, 'moves_new2':0, 'moves_new3':0}) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "## Feature Engineering and Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:47:21.306662Z",
     "start_time": "2019-10-30T08:47:21.291569Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['rated',\n",
       " 'turns',\n",
       " 'victory_status',\n",
       " 'winner',\n",
       " 'white_rating',\n",
       " 'black_rating',\n",
       " 'opening_name',\n",
       " 'opening_ply',\n",
       " 'increment1',\n",
       " 'increment2',\n",
       " 'moves_new0',\n",
       " 'moves_new1',\n",
       " 'moves_new2',\n",
       " 'moves_new3']"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#let's see what columns we are left with\n",
    "df.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We create here a pipeline to preprocess our columns: on categorical features we apply StringIndexer and OHE, whereas we apply only StringIndexer on the label column; finally, we apply VectorAssembler on all the features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:47:24.491888Z",
     "start_time": "2019-10-30T08:47:24.360137Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "# Spark Pipeline\n",
    "\n",
    "#categorical features\n",
    "cat_features = ['rated', 'opening_name'] + [\"moves_new{0}\".format(i) for i in range(4)]\n",
    "\n",
    "#numerical_features\n",
    "num_features = ['turns', 'white_rating', 'black_rating', 'opening_ply', 'increment1', 'increment2']\n",
    "\n",
    "#label to predict\n",
    "label = 'winner'\n",
    "\n",
    "# Pipeline Stages List\n",
    "stages = []\n",
    "\n",
    "# Loop for StringIndexer and OHE for Categorical Variables\n",
    "for features in cat_features:\n",
    "    \n",
    "    # Index Categorical Features\n",
    "    string_indexer = StringIndexer(inputCol=features, outputCol=features + \"_index\")\n",
    "    \n",
    "    # One Hot Encode Categorical Features\n",
    "    encoder = OneHotEncoderEstimator(inputCols=[string_indexer.getOutputCol()],\n",
    "                                     outputCols=[features + \"_class_vec\"])\n",
    "    # Append Pipeline Stages\n",
    "    stages += [string_indexer, encoder]\n",
    "    \n",
    "# Index Label Feature\n",
    "label_str_index =  StringIndexer(inputCol=label, outputCol=\"label_index\")\n",
    "\n",
    "# Assemble or Concat the Categorical Features and Numeric Features\n",
    "assembler_inputs = [feature + \"_class_vec\" for feature in cat_features] + num_features\n",
    "\n",
    "assembler = VectorAssembler(inputCols=assembler_inputs, outputCol=\"features\") \n",
    "\n",
    "stages += [label_str_index, assembler]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Here we see the details of the stages**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:47:26.563438Z",
     "start_time": "2019-10-30T08:47:26.556215Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[StringIndexer_844b3ff70bd8,\n",
       " OneHotEncoderEstimator_80efe4b2af3a,\n",
       " StringIndexer_d7607f841299,\n",
       " OneHotEncoderEstimator_85f8980c3412,\n",
       " StringIndexer_6aab36eb1dad,\n",
       " OneHotEncoderEstimator_694757b6bea3,\n",
       " StringIndexer_17d4c89bb983,\n",
       " OneHotEncoderEstimator_6b4aca6b8aa9,\n",
       " StringIndexer_b6411b8ad663,\n",
       " OneHotEncoderEstimator_facc55adf211,\n",
       " StringIndexer_efb0c90d30bf,\n",
       " OneHotEncoderEstimator_e8a3400e0ce7,\n",
       " StringIndexer_ce07482d3ac5,\n",
       " VectorAssembler_a2b52d3ebab3]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stages"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Fitting the pipeline and transforming the dataframe**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:49:53.788158Z",
     "start_time": "2019-10-30T08:47:29.755574Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "pipeline initialized\n",
      "model fitted\n"
     ]
    }
   ],
   "source": [
    "# Set Pipeline\n",
    "pipeline = Pipeline(stages=stages)\n",
    "print(\"pipeline initialized\")\n",
    "\n",
    "# Fit Pipeline to Data\n",
    "pipeline_model = pipeline.fit(df)\n",
    "print(\"model fitted\")\n",
    "\n",
    "# Transform Data using Fitted Pipeline\n",
    "df_transform = pipeline_model.transform(df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see the transformed datasets**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:50:15.451984Z",
     "start_time": "2019-10-30T08:49:53.792549Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "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>rated</th>\n",
       "      <th>turns</th>\n",
       "      <th>victory_status</th>\n",
       "      <th>winner</th>\n",
       "      <th>white_rating</th>\n",
       "      <th>black_rating</th>\n",
       "      <th>opening_name</th>\n",
       "      <th>opening_ply</th>\n",
       "      <th>increment1</th>\n",
       "      <th>increment2</th>\n",
       "      <th>...</th>\n",
       "      <th>moves_new0_index</th>\n",
       "      <th>moves_new0_class_vec</th>\n",
       "      <th>moves_new1_index</th>\n",
       "      <th>moves_new1_class_vec</th>\n",
       "      <th>moves_new2_index</th>\n",
       "      <th>moves_new2_class_vec</th>\n",
       "      <th>moves_new3_index</th>\n",
       "      <th>moves_new3_class_vec</th>\n",
       "      <th>label_index</th>\n",
       "      <th>features</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>True</td>\n",
       "      <td>139</td>\n",
       "      <td>draw</td>\n",
       "      <td>draw</td>\n",
       "      <td>1736</td>\n",
       "      <td>1893</td>\n",
       "      <td>Old Benoni Defense</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>(0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>337.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>110.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>True</td>\n",
       "      <td>96</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>2209</td>\n",
       "      <td>2044</td>\n",
       "      <td>Sicilian Defense: Alapin Variation</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>2.0</td>\n",
       "      <td>(0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>97.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>114.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>True</td>\n",
       "      <td>58</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>1652</td>\n",
       "      <td>1624</td>\n",
       "      <td>Scotch Game: Malaniuk Variation</td>\n",
       "      <td>8</td>\n",
       "      <td>6</td>\n",
       "      <td>8</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>83.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>321.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>False</td>\n",
       "      <td>39</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>1812</td>\n",
       "      <td>1820</td>\n",
       "      <td>Queen's Pawn Game: London System</td>\n",
       "      <td>5</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>1.0</td>\n",
       "      <td>(0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>52.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>578.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>True</td>\n",
       "      <td>93</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>2217</td>\n",
       "      <td>1970</td>\n",
       "      <td>King's Pawn Game: Nimzowitsch Defense</td>\n",
       "      <td>3</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>7.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, ...</td>\n",
       "      <td>403.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>224.0</td>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 28 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   rated  turns victory_status winner  white_rating  black_rating  \\\n",
       "0   True    139           draw   draw          1736          1893   \n",
       "1   True     96         resign  black          2209          2044   \n",
       "2   True     58         resign  black          1652          1624   \n",
       "3  False     39         resign  white          1812          1820   \n",
       "4   True     93         resign  white          2217          1970   \n",
       "\n",
       "                            opening_name  opening_ply  increment1  increment2  \\\n",
       "0                     Old Benoni Defense            2          10           0   \n",
       "1     Sicilian Defense: Alapin Variation            3           8           0   \n",
       "2        Scotch Game: Malaniuk Variation            8           6           8   \n",
       "3       Queen's Pawn Game: London System            5          10           0   \n",
       "4  King's Pawn Game: Nimzowitsch Defense            3          10           0   \n",
       "\n",
       "   ... moves_new0_index                               moves_new0_class_vec  \\\n",
       "0  ...              1.0  (0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "1  ...              0.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "2  ...              0.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "3  ...              1.0  (0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "4  ...              0.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "\n",
       "  moves_new1_index                               moves_new1_class_vec  \\\n",
       "0              2.0  (0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "1              2.0  (0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "2              0.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "3              1.0  (0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "4              7.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, ...   \n",
       "\n",
       "   moves_new2_index                               moves_new2_class_vec  \\\n",
       "0             337.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "1              97.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "2              83.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "3              52.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "4             403.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "\n",
       "   moves_new3_index                               moves_new3_class_vec  \\\n",
       "0             110.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "1             114.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "2             321.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "3             578.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "4             224.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...   \n",
       "\n",
       "   label_index                                           features  \n",
       "0          2.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "1          1.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "2          1.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "3          0.0  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "4          0.0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...  \n",
       "\n",
       "[5 rows x 28 columns]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Preview Newly Transformed Data\n",
    "df_transform.limit(5).toPandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:50:34.844562Z",
     "start_time": "2019-10-30T08:50:15.457708Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>features</th>\n",
       "      <th>label_index</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>2.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            features  label_index\n",
       "0  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...          2.0\n",
       "1  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...          1.0\n",
       "2  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...          1.0\n",
       "3  (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...          0.0\n",
       "4  (1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, ...          0.0"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Select only 'features' and 'label_index' for Final Dataframe\n",
    "df_transform_fin = df_transform.select('features','label_index')\n",
    "df_transform_fin.limit(5).toPandas()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's split the dataset to train models**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:50:34.875564Z",
     "start_time": "2019-10-30T08:50:34.850249Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "# Split Data into Train / Test Sets\n",
    "train_data, test_data = df_transform_fin.randomSplit([.8, .2],seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:50:53.757356Z",
     "start_time": "2019-10-30T08:50:34.879326Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+-----------+\n",
      "|            features|label_index|\n",
      "+--------------------+-----------+\n",
      "|(4952,[0,3,1477,1...|        0.0|\n",
      "|(4952,[0,9,1477,1...|        1.0|\n",
      "|(4952,[0,9,1477,1...|        1.0|\n",
      "|(4952,[0,9,1477,1...|        1.0|\n",
      "|(4952,[0,10,1478,...|        0.0|\n",
      "+--------------------+-----------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "train_data.show(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Definition, Training and Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Machine Learning Models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Decision Tree"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's first try with a simple Decision Tree, using 5 as maxDepth and Gini as impurity criterion**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:25:00.052666Z",
     "start_time": "2019-10-29T08:22:53.943209Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----------+--------------------+----------+--------------------+\n",
      "|label_index|       rawPrediction|prediction|         probability|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "|        0.0|[1375.0,1715.0,16...|       1.0|[0.42268675069166...|\n",
      "|        1.0|   [54.0,135.0,12.0]|       1.0|[0.26865671641791...|\n",
      "|        0.0|[2392.0,1734.0,18...|       0.0|[0.55473098330241...|\n",
      "|        0.0|   [359.0,78.0,10.0]|       0.0|[0.80313199105145...|\n",
      "|        0.0|[2392.0,1734.0,18...|       0.0|[0.55473098330241...|\n",
      "|        0.0|[1375.0,1715.0,16...|       1.0|[0.42268675069166...|\n",
      "|        2.0|[1375.0,1715.0,16...|       1.0|[0.42268675069166...|\n",
      "|        0.0|  [966.0,361.0,79.0]|       0.0|[0.68705547652916...|\n",
      "|        1.0|    [60.0,433.0,9.0]|       1.0|[0.11952191235059...|\n",
      "|        0.0|[2392.0,1734.0,18...|       0.0|[0.55473098330241...|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dt = DecisionTreeClassifier(featuresCol = 'features', labelCol = 'label_index',\n",
    "                            maxDepth = 5, seed=42)\n",
    "dtModel = dt.fit(train_data)\n",
    "\n",
    "predictions = dtModel.transform(test_data)\n",
    "predictions.select('label_index', 'rawPrediction', 'prediction', 'probability').show(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Performances are not so bad, but there is room for improvement, considering that the model hasn't even been tuned**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:25:23.099396Z",
     "start_time": "2019-10-29T08:25:00.061496Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy = 0.59337\n"
     ]
    }
   ],
   "source": [
    "evaluator = MulticlassClassificationEvaluator(\n",
    "    labelCol=\"label_index\", predictionCol=\"prediction\", metricName=\"accuracy\")\n",
    "accuracy = evaluator.evaluate(predictions)\n",
    "print(\"Test Set Accuracy = %g\" % (accuracy))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's try with a Random Forest, using 20 trees and 5 as maxDepth**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:31:45.947459Z",
     "start_time": "2019-10-29T08:26:59.523477Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----------+--------------------+----------+--------------------+\n",
      "|label_index|       rawPrediction|prediction|         probability|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "|        0.0|[9.92118618275944...|       0.0|[0.49605930913797...|\n",
      "|        1.0|[10.1693235631783...|       0.0|[0.50846617815891...|\n",
      "|        0.0|[9.92118618275944...|       0.0|[0.49605930913797...|\n",
      "|        0.0|[10.1693235631783...|       0.0|[0.50846617815891...|\n",
      "|        0.0|[10.0308165621266...|       0.0|[0.50154082810633...|\n",
      "|        0.0|[9.92118618275944...|       0.0|[0.49605930913797...|\n",
      "|        2.0|[9.85599977960717...|       0.0|[0.49279998898035...|\n",
      "|        0.0|[9.92118618275944...|       0.0|[0.49605930913797...|\n",
      "|        1.0|[9.92118618275944...|       0.0|[0.49605930913797...|\n",
      "|        0.0|[10.2961729806778...|       0.0|[0.51480864903389...|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dt2 = RandomForestClassifier(featuresCol = 'features', labelCol = 'label_index', numTrees=20,\n",
    "                             maxDepth = 5, seed=42)\n",
    "dt2Model = dt2.fit(train_data)\n",
    "\n",
    "predictions2 = dt2Model.transform(test_data)\n",
    "predictions2.select('label_index', 'rawPrediction', 'prediction', 'probability').show(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**The accuracy level is lower than before, this could be due to the little number of trees used**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:32:07.673827Z",
     "start_time": "2019-10-29T08:31:45.982148Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy = 0.548822\n"
     ]
    }
   ],
   "source": [
    "evaluator2 = MulticlassClassificationEvaluator(\n",
    "    labelCol=\"label_index\", predictionCol=\"prediction\", metricName=\"accuracy\")\n",
    "accuracy2 = evaluator2.evaluate(predictions2)\n",
    "print(\"Test Set Accuracy = %g\" % (accuracy2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's inspect the importance of every single feature in the classification task. The top15 are shown below, along with their importance score**\n",
    "\n",
    "**We see that, of course, the last moves are very important, along with the black player rating**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T10:29:30.013107Z",
     "start_time": "2019-10-29T10:29:29.839549Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature Importance according to RandomForestClassifier\n"
     ]
    },
    {
     "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>idx</th>\n",
       "      <th>name</th>\n",
       "      <th>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2900</th>\n",
       "      <td>2894</td>\n",
       "      <td>moves_new3_class_vec_Qxf2#</td>\n",
       "      <td>0.086035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1522</th>\n",
       "      <td>1516</td>\n",
       "      <td>moves_new2_class_vec_Kh8</td>\n",
       "      <td>0.061678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2901</th>\n",
       "      <td>2895</td>\n",
       "      <td>moves_new3_class_vec_Qf2#</td>\n",
       "      <td>0.050262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1526</th>\n",
       "      <td>1520</td>\n",
       "      <td>moves_new2_class_vec_Kg1</td>\n",
       "      <td>0.045928</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1533</th>\n",
       "      <td>1527</td>\n",
       "      <td>moves_new2_class_vec_Ke2</td>\n",
       "      <td>0.042275</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4946</td>\n",
       "      <td>black_rating</td>\n",
       "      <td>0.036453</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1546</th>\n",
       "      <td>1540</td>\n",
       "      <td>moves_new2_class_vec_Kd2</td>\n",
       "      <td>0.035300</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2898</th>\n",
       "      <td>2892</td>\n",
       "      <td>moves_new3_class_vec_Qf7#</td>\n",
       "      <td>0.035298</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1543</th>\n",
       "      <td>1537</td>\n",
       "      <td>moves_new2_class_vec_Kd1</td>\n",
       "      <td>0.031475</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1536</th>\n",
       "      <td>1530</td>\n",
       "      <td>moves_new2_class_vec_Ke1</td>\n",
       "      <td>0.031467</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1540</th>\n",
       "      <td>1534</td>\n",
       "      <td>moves_new2_class_vec_Kf2</td>\n",
       "      <td>0.030360</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1597</th>\n",
       "      <td>1591</td>\n",
       "      <td>moves_new2_class_vec_Rf1</td>\n",
       "      <td>0.025012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1523</th>\n",
       "      <td>1517</td>\n",
       "      <td>moves_new2_class_vec_Kf8</td>\n",
       "      <td>0.022163</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1528</th>\n",
       "      <td>1522</td>\n",
       "      <td>moves_new2_class_vec_Kg8</td>\n",
       "      <td>0.020249</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1529</th>\n",
       "      <td>1523</td>\n",
       "      <td>moves_new2_class_vec_Kd8</td>\n",
       "      <td>0.019544</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       idx                        name     score\n",
       "2900  2894  moves_new3_class_vec_Qxf2#  0.086035\n",
       "1522  1516    moves_new2_class_vec_Kh8  0.061678\n",
       "2901  2895   moves_new3_class_vec_Qf2#  0.050262\n",
       "1526  1520    moves_new2_class_vec_Kg1  0.045928\n",
       "1533  1527    moves_new2_class_vec_Ke2  0.042275\n",
       "0     4946                black_rating  0.036453\n",
       "1546  1540    moves_new2_class_vec_Kd2  0.035300\n",
       "2898  2892   moves_new3_class_vec_Qf7#  0.035298\n",
       "1543  1537    moves_new2_class_vec_Kd1  0.031475\n",
       "1536  1530    moves_new2_class_vec_Ke1  0.031467\n",
       "1540  1534    moves_new2_class_vec_Kf2  0.030360\n",
       "1597  1591    moves_new2_class_vec_Rf1  0.025012\n",
       "1523  1517    moves_new2_class_vec_Kf8  0.022163\n",
       "1528  1522    moves_new2_class_vec_Kg8  0.020249\n",
       "1529  1523    moves_new2_class_vec_Kd8  0.019544"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#custom function to plot feature importance as dataframe\n",
    "def ExtractFeatureImp(featureImp, dataset, featuresCol):\n",
    "    list_extract = []\n",
    "    for i in dataset.schema[featuresCol].metadata[\"ml_attr\"][\"attrs\"]:\n",
    "        list_extract = list_extract + dataset.schema[featuresCol].metadata[\"ml_attr\"][\"attrs\"][i]\n",
    "    varlist = pd.DataFrame(list_extract)\n",
    "    varlist['score'] = varlist['idx'].apply(lambda x: featureImp[x])\n",
    "    return(varlist.sort_values('score', ascending = False))\n",
    "\n",
    "print(\"Feature Importance according to RandomForestClassifier\")\n",
    "ExtractFeatureImp(dt2Model.featureImportances, df_transform_fin, \"features\").head(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's increase the number of trees and compare the obtained result with the previous one**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T15:51:02.225659Z",
     "start_time": "2019-10-29T15:34:56.798741Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----------+--------------------+----------+--------------------+\n",
      "|label_index|       rawPrediction|prediction|         probability|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "|        0.0|[24.9062157219945...|       0.0|[0.49812431443989...|\n",
      "|        1.0|[24.3726724937313...|       0.0|[0.48745344987462...|\n",
      "|        0.0|[25.1859043151886...|       0.0|[0.50371808630377...|\n",
      "|        0.0|[25.0519876756169...|       0.0|[0.50103975351233...|\n",
      "|        0.0|[26.7405012631316...|       0.0|[0.53481002526263...|\n",
      "|        0.0|[25.0480463304278...|       0.0|[0.50096092660855...|\n",
      "|        2.0|[24.5780693632626...|       0.0|[0.49156138726525...|\n",
      "|        0.0|[25.0205572718997...|       0.0|[0.50041114543799...|\n",
      "|        1.0|[24.5687948063188...|       0.0|[0.49137589612637...|\n",
      "|        0.0|[26.8260392005576...|       0.0|[0.53652078401115...|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dt2_bis = RandomForestClassifier(featuresCol = 'features', labelCol = 'label_index', numTrees=50,\n",
    "                             maxDepth = 7, seed=42)\n",
    "dt2bModel = dt2_bis.fit(train_data)\n",
    "\n",
    "predictions2b = dt2bModel.transform(test_data)\n",
    "predictions2b.select('label_index', 'rawPrediction', 'prediction', 'probability').show(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**The improvement is slight, so new alternatives are to be preferred over this model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T15:51:25.170222Z",
     "start_time": "2019-10-29T15:51:02.262636Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy = 0.570319\n"
     ]
    }
   ],
   "source": [
    "evaluator2b = MulticlassClassificationEvaluator(\n",
    "    labelCol=\"label_index\", predictionCol=\"prediction\", metricName=\"accuracy\")\n",
    "accuracy2b = evaluator2b.evaluate(predictions2b)\n",
    "print(\"Test Set Accuracy = %g\" % (accuracy2b))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's use the famous logistic regression, in the multiclass version**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:38:41.817926Z",
     "start_time": "2019-10-29T08:37:22.366582Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----------+--------------------+----------+--------------------+\n",
      "|label_index|       rawPrediction|prediction|         probability|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "|        0.0|[0.07832550896203...|       1.0|[0.35426958643812...|\n",
      "|        1.0|[-0.7747465818665...|       1.0|[0.01446011173150...|\n",
      "|        0.0|[2.32551393344361...|       0.0|[0.91492911171162...|\n",
      "|        0.0|[2.44407460467075...|       0.0|[0.79324394198562...|\n",
      "|        0.0|[6.27017707591990...|       0.0|[0.99983544741904...|\n",
      "|        0.0|[-2.1241437294911...|       1.0|[3.04282149222622...|\n",
      "|        2.0|[1.82497119028347...|       0.0|[0.69026548653943...|\n",
      "|        0.0|[2.20055334692985...|       0.0|[0.92918447626717...|\n",
      "|        1.0|[1.14902753173228...|       1.0|[0.10971668049453...|\n",
      "|        0.0|[0.68414187201709...|       1.0|[0.36142987688250...|\n",
      "+-----------+--------------------+----------+--------------------+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dt3 = LogisticRegression(featuresCol = 'features', labelCol = 'label_index', maxIter = 10)\n",
    "\n",
    "dt3Model = dt3.fit(train_data)\n",
    "\n",
    "predictions3 = dt3Model.transform(test_data)\n",
    "predictions3.select('label_index', 'rawPrediction', 'prediction', 'probability').show(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Without having been tuned, this model considerably outperforms the preceding ones, with an astonishing 70% accuracy on the test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:39:06.708178Z",
     "start_time": "2019-10-29T08:38:41.828405Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Set Accuracy = 0.699042\n"
     ]
    }
   ],
   "source": [
    "evaluator3 = MulticlassClassificationEvaluator(\n",
    "    labelCol=\"label_index\", predictionCol=\"prediction\", metricName=\"accuracy\")\n",
    "accuracy3 = evaluator3.evaluate(predictions3)\n",
    "print(\"Test Set Accuracy = %g\" % (accuracy3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Given the promising results, let's implement a GridSearch with CrossValidation in order to further improve the model**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Here we define the parameters that we want to tune using a GridSearch approach**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T08:51:54.642499Z",
     "start_time": "2019-10-29T08:51:54.622211Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#paramGrid = ParamGridBuilder()\\\n",
    "#    .addGrid(dt3.aggregationDepth,[2,5,10])\\\n",
    "#    .addGrid(dt3.elasticNetParam,[0.0, 0.5, 1.0])\\\n",
    "#    .addGrid(dt3.fitIntercept,[False, True])\\\n",
    "#    .addGrid(dt3.maxIter,[10, 100, 1000])\\\n",
    "#    .addGrid(dt3.regParam,[0.01, 0.5, 2.0]) \\\n",
    "#    .build()\n",
    "\n",
    "paramGrid = ParamGridBuilder()\\\n",
    "    .addGrid(dt3.aggregationDepth,[2,5])\\\n",
    "    .addGrid(dt3.elasticNetParam,[0.0, 0.5, 1.0])\\\n",
    "    .addGrid(dt3.regParam,[0.01, 0.5, 2.0]) \\\n",
    "    .build()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We now perform the GridSearch with CrossValidation**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T10:13:01.074781Z",
     "start_time": "2019-10-29T08:52:00.585099Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception ignored in: <object repr() failed>\n",
      "Traceback (most recent call last):\n",
      "  File \"/usr/local/Cellar/apache-spark/2.4.3/libexec/python/pyspark/ml/wrapper.py\", line 40, in __del__\n",
      "    if SparkContext._active_spark_context and self._java_obj is not None:\n",
      "AttributeError: 'LogisticRegression' object has no attribute '_java_obj'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model fitted\n",
      "The accuracy on the training set after cv is 0.7899291896144768\n",
      "The accuracy on the test set after cv is 0.7311577311577312\n"
     ]
    }
   ],
   "source": [
    "#it took 1h20min\n",
    "\n",
    "#Create 3-fold CrossValidator\n",
    "cv = CrossValidator(estimator=dt3, estimatorParamMaps=paramGrid, evaluator=evaluator3, numFolds=3)\n",
    "\n",
    "# Run cross validations\n",
    "cvModel = cv.fit(train_data)\n",
    "print(\"Model fitted\")\n",
    "\n",
    "\n",
    "predict_train=cvModel.transform(train_data)\n",
    "predict_test=cvModel.transform(test_data)\n",
    "print(\"The accuracy on the training set after cv is {}\".format(evaluator3.evaluate(predict_train)))\n",
    "print(\"The accuracy on the test set after cv is {}\".format(evaluator3.evaluate(predict_test)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**A further improvement of more than 3% has been obtained**\n",
    "\n",
    "**Let's see the best parameters that contributed to this accuracy score**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T10:19:28.208830Z",
     "start_time": "2019-10-29T10:19:28.201744Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "bestmod = cvModel.bestModel\n",
    "bestParams = bestmod.extractParamMap()\n",
    "# best are aggregationDepth=2, elasticNetParam=0.5, regParam=0.01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T10:19:29.599720Z",
     "start_time": "2019-10-29T10:19:29.593294Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='LogisticRegression_38de24403f56', name='aggregationDepth', doc='suggested depth for treeAggregate (>= 2)'): 2,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='elasticNetParam', doc='the ElasticNet mixing parameter, in range [0, 1]. For alpha = 0, the penalty is an L2 penalty. For alpha = 1, it is an L1 penalty'): 0.5,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='family', doc='The name of family which is a description of the label distribution to be used in the model. Supported options: auto, binomial, multinomial.'): 'auto',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='featuresCol', doc='features column name'): 'features',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='fitIntercept', doc='whether to fit an intercept term'): True,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='labelCol', doc='label column name'): 'label_index',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='maxIter', doc='maximum number of iterations (>= 0)'): 10,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='predictionCol', doc='prediction column name'): 'prediction',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities'): 'probability',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name'): 'rawPrediction',\n",
       " Param(parent='LogisticRegression_38de24403f56', name='regParam', doc='regularization parameter (>= 0)'): 0.01,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='standardization', doc='whether to standardize the training features before fitting the model'): True,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='threshold', doc='threshold in binary classification prediction, in range [0, 1]'): 0.5,\n",
       " Param(parent='LogisticRegression_38de24403f56', name='tol', doc='the convergence tolerance for iterative algorithms (>= 0)'): 1e-06}"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bestParams"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's try again to get something more out the GridSearch, using the parameters found earlier and trying to tune the others (we split the GridSearch due to time constraints, but it should have been done in one single step)**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T08:51:20.529200Z",
     "start_time": "2019-10-30T08:51:20.523758Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "paramGrid2 = ParamGridBuilder()\\\n",
    "    .addGrid(dt3.aggregationDepth,[2])\\\n",
    "    .addGrid(dt3.elasticNetParam,[0.5])\\\n",
    "    .addGrid(dt3.fitIntercept,[False, True])\\\n",
    "    .addGrid(dt3.maxIter,[10, 100])\\\n",
    "    .addGrid(dt3.regParam,[0.01]) \\\n",
    "    .build()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We manage to further improve our model, reaching an accuracy of 74% on the test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T09:35:13.455642Z",
     "start_time": "2019-10-30T08:51:32.663175Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model fitted\n",
      "The accuracy on the training set after cv is 0.7934041437188566\n",
      "The accuracy on the test set after cv is 0.7402227402227403\n"
     ]
    }
   ],
   "source": [
    "#it took 43min\n",
    "\n",
    "#Create 3-fold CrossValidator\n",
    "cv2 = CrossValidator(estimator=dt3, estimatorParamMaps=paramGrid2, evaluator=evaluator3, numFolds=3)\n",
    "\n",
    "# Run cross validations\n",
    "cvModel2 = cv2.fit(train_data)\n",
    "print(\"Model fitted\")\n",
    "\n",
    "\n",
    "predict_train2=cvModel2.transform(train_data)\n",
    "predict_test2=cvModel2.transform(test_data)\n",
    "print(\"The accuracy on the training set after cv is {}\".format(evaluator3.evaluate(predict_train2)))\n",
    "print(\"The accuracy on the test set after cv is {}\".format(evaluator3.evaluate(predict_test2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T09:53:17.891159Z",
     "start_time": "2019-10-30T09:53:17.884660Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "bestmod2 = cvModel2.bestModel\n",
    "bestParams2 = bestmod2.extractParamMap()\n",
    "# best are aggregationDepth=2, elasticNetParam=0.5, regParam=0.01\n",
    "# new best are fit_intercept=False, maxIter=100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T09:53:20.382701Z",
     "start_time": "2019-10-30T09:53:20.374751Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='LogisticRegression_50a11c3e6bc3', name='aggregationDepth', doc='suggested depth for treeAggregate (>= 2)'): 2,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='elasticNetParam', doc='the ElasticNet mixing parameter, in range [0, 1]. For alpha = 0, the penalty is an L2 penalty. For alpha = 1, it is an L1 penalty'): 0.5,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='family', doc='The name of family which is a description of the label distribution to be used in the model. Supported options: auto, binomial, multinomial.'): 'auto',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='featuresCol', doc='features column name'): 'features',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='fitIntercept', doc='whether to fit an intercept term'): False,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='labelCol', doc='label column name'): 'label_index',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='maxIter', doc='maximum number of iterations (>= 0)'): 100,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='predictionCol', doc='prediction column name'): 'prediction',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities'): 'probability',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name'): 'rawPrediction',\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='regParam', doc='regularization parameter (>= 0)'): 0.01,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='standardization', doc='whether to standardize the training features before fitting the model'): True,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='threshold', doc='threshold in binary classification prediction, in range [0, 1]'): 0.5,\n",
       " Param(parent='LogisticRegression_50a11c3e6bc3', name='tol', doc='the convergence tolerance for iterative algorithms (>= 0)'): 1e-06}"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bestParams2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deep Learning Models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Neural Network without scaling numerical features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Unfortunately we will implement and run some deep neural nets using Pandas because SystemML continued to create problems with Python3, and also switching to Python2 wasn't successful because of other incompatibilities**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's assign the number of classes, in our case 3 (white winner, black winner, draw), and the input dimension, the number of feature columns after all the transformations**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-30T11:03:51.096491Z",
     "start_time": "2019-10-30T11:03:51.090232Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of classes: 3\n",
      "Input dimension: 4956\n"
     ]
    }
   ],
   "source": [
    "# Number of Classes\n",
    "nb_classes = 3  #train_data.select(\"label_index\").distinct().count()\n",
    "print(\"Number of classes:\", nb_classes)\n",
    "\n",
    "# Number of Inputs or Input Dimensions\n",
    "input_dim = 4956 #len(train_data.select(\"features\").first()[0])  #4953\n",
    "print(\"Input dimension:\", input_dim)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**After many trials and errors, the following architecture has proven to be the best in predicting the label.**\n",
    "\n",
    "**It consists of 5 layers with decreasing number of neurons, 2 l2-regularizers and 5 dropout layers with 0.2 dropout-rate.\n",
    "The activation functions are all ReLu except for the last one, which is a Softmax, the right choice for multiclass classification**\n",
    "\n",
    "**I chose Adam as optimizer as a result of trial and error processes, and also because it combines the best properties of the AdaGrad and RMSProp algorithms to provide an optimization algorithm that can handle sparse gradients on noisy problems.**\n",
    "\n",
    "**Moreover, I chose Accuracy as evaluation metric because the Draw class is very tiny and because it is the most immediate to understand. I experimented also with F1-score and precision, but the information gain was very slight, so I sticked to the basic accuracy**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:19:45.026842Z",
     "start_time": "2019-10-29T11:19:44.610303Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Set up Deep Learning Model / Architecture\n",
    "model = Sequential()\n",
    "model.add(Dense(512, input_shape=(input_dim,), activity_regularizer=regularizers.l2(0.01)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(rate=0.2))\n",
    "model.add(Dense(256, activity_regularizer=regularizers.l2(0.01)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(rate=0.2))\n",
    "model.add(Dense(256))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(rate=0.2))\n",
    "model.add(Dense(128))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(rate=0.2))\n",
    "model.add(Dense(nb_classes))\n",
    "model.add(Activation('softmax'))\n",
    "model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's have a look at the model structure**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:19:48.354025Z",
     "start_time": "2019-10-29T11:19:48.339851Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_46 (Dense)             (None, 512)               2537984   \n",
      "_________________________________________________________________\n",
      "activation_46 (Activation)   (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_37 (Dropout)         (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_47 (Dense)             (None, 256)               131328    \n",
      "_________________________________________________________________\n",
      "activation_47 (Activation)   (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dropout_38 (Dropout)         (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_48 (Dense)             (None, 256)               65792     \n",
      "_________________________________________________________________\n",
      "activation_48 (Activation)   (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dropout_39 (Dropout)         (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_49 (Dense)             (None, 128)               32896     \n",
      "_________________________________________________________________\n",
      "activation_49 (Activation)   (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_40 (Dropout)         (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_50 (Dense)             (None, 3)                 387       \n",
      "_________________________________________________________________\n",
      "activation_50 (Activation)   (None, 3)                 0         \n",
      "=================================================================\n",
      "Total params: 2,768,387\n",
      "Trainable params: 2,768,387\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We convert the Spark dataframe to a Pandas dataframe, transform the label column with a LabelEncoder, then we drop irrelevant columns as well as the victory_status column, which would falsify the prediction; then we perform OHE. We finally split the dataframe into training set and test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:19:59.948269Z",
     "start_time": "2019-10-29T11:19:55.210766Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "le = LabelEncoder()\n",
    "\n",
    "dfp = df.toPandas()\n",
    "col_y = le.fit_transform(dfp.winner)\n",
    "dfp2 = dfp.drop([\"victory_status\", \"winner\", \"rated\"], 1)\n",
    "dfp3 = pd.get_dummies(dfp2)\n",
    "dfp3[\"winner\"] = col_y\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(dfp3.drop([\"winner\"],1), \n",
    "                                                    dfp3.winner, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Just to see the correspondence between numbers and labels**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:20:02.907706Z",
     "start_time": "2019-10-29T11:20:02.900093Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['black', 'draw', 'white'], dtype=object)"
      ]
     },
     "execution_count": 144,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "le.inverse_transform([0, 1, 2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's have a look at the final dataframe that will be fed to the net**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:20:04.821683Z",
     "start_time": "2019-10-29T11:20:04.786477Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>rated</th>\n",
       "      <th>turns</th>\n",
       "      <th>victory_status</th>\n",
       "      <th>winner</th>\n",
       "      <th>white_rating</th>\n",
       "      <th>black_rating</th>\n",
       "      <th>opening_name</th>\n",
       "      <th>opening_ply</th>\n",
       "      <th>increment1</th>\n",
       "      <th>increment2</th>\n",
       "      <th>moves_new0</th>\n",
       "      <th>moves_new1</th>\n",
       "      <th>moves_new2</th>\n",
       "      <th>moves_new3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>True</td>\n",
       "      <td>139</td>\n",
       "      <td>draw</td>\n",
       "      <td>draw</td>\n",
       "      <td>1736</td>\n",
       "      <td>1893</td>\n",
       "      <td>Old Benoni Defense</td>\n",
       "      <td>2</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>d4</td>\n",
       "      <td>c5</td>\n",
       "      <td>Rh3</td>\n",
       "      <td>Kd2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>True</td>\n",
       "      <td>96</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>2209</td>\n",
       "      <td>2044</td>\n",
       "      <td>Sicilian Defense: Alapin Variation</td>\n",
       "      <td>3</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>e4</td>\n",
       "      <td>c5</td>\n",
       "      <td>Re2</td>\n",
       "      <td>a3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>True</td>\n",
       "      <td>58</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>1652</td>\n",
       "      <td>1624</td>\n",
       "      <td>Scotch Game: Malaniuk Variation</td>\n",
       "      <td>8</td>\n",
       "      <td>6</td>\n",
       "      <td>8</td>\n",
       "      <td>e4</td>\n",
       "      <td>e5</td>\n",
       "      <td>h3</td>\n",
       "      <td>Bd5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>False</td>\n",
       "      <td>39</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>1812</td>\n",
       "      <td>1820</td>\n",
       "      <td>Queen's Pawn Game: London System</td>\n",
       "      <td>5</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>d4</td>\n",
       "      <td>d5</td>\n",
       "      <td>h6</td>\n",
       "      <td>Qc4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>True</td>\n",
       "      <td>93</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>2217</td>\n",
       "      <td>1970</td>\n",
       "      <td>King's Pawn Game: Nimzowitsch Defense</td>\n",
       "      <td>3</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>e4</td>\n",
       "      <td>Nc6</td>\n",
       "      <td>Rd2+</td>\n",
       "      <td>Bxd2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   rated  turns victory_status winner  white_rating  black_rating  \\\n",
       "0   True    139           draw   draw          1736          1893   \n",
       "1   True     96         resign  black          2209          2044   \n",
       "2   True     58         resign  black          1652          1624   \n",
       "3  False     39         resign  white          1812          1820   \n",
       "4   True     93         resign  white          2217          1970   \n",
       "\n",
       "                            opening_name  opening_ply  increment1  increment2  \\\n",
       "0                     Old Benoni Defense            2          10           0   \n",
       "1     Sicilian Defense: Alapin Variation            3           8           0   \n",
       "2        Scotch Game: Malaniuk Variation            8           6           8   \n",
       "3       Queen's Pawn Game: London System            5          10           0   \n",
       "4  King's Pawn Game: Nimzowitsch Defense            3          10           0   \n",
       "\n",
       "  moves_new0 moves_new1 moves_new2 moves_new3  \n",
       "0         d4         c5        Rh3        Kd2  \n",
       "1         e4         c5        Re2         a3  \n",
       "2         e4         e5         h3        Bd5  \n",
       "3         d4         d5         h6        Qc4  \n",
       "4         e4        Nc6       Rd2+       Bxd2  "
      ]
     },
     "execution_count": 145,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dfp.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Finally, we train the net for 30 epochs, using a batch size of 256**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:23:51.618672Z",
     "start_time": "2019-10-29T11:20:08.037909Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15290 samples, validate on 3823 samples\n",
      "Epoch 1/30\n",
      "15290/15290 [==============================] - 10s 645us/step - loss: 474516.2953 - acc: 0.5012 - val_loss: 16100.8306 - val_acc: 0.5177\n",
      "Epoch 2/30\n",
      "15290/15290 [==============================] - 7s 468us/step - loss: 11158.7465 - acc: 0.4976 - val_loss: 8310.4972 - val_acc: 0.5443\n",
      "Epoch 3/30\n",
      "15290/15290 [==============================] - 7s 434us/step - loss: 6999.5011 - acc: 0.5167 - val_loss: 5511.2367 - val_acc: 0.5582\n",
      "Epoch 4/30\n",
      "15290/15290 [==============================] - 7s 440us/step - loss: 4642.3637 - acc: 0.5315 - val_loss: 3668.3742 - val_acc: 0.5739\n",
      "Epoch 5/30\n",
      "15290/15290 [==============================] - 7s 442us/step - loss: 3111.8474 - acc: 0.5519 - val_loss: 2472.1366 - val_acc: 0.5749\n",
      "Epoch 6/30\n",
      "15290/15290 [==============================] - 7s 447us/step - loss: 2102.4305 - acc: 0.5712 - val_loss: 1688.1430 - val_acc: 0.6066\n",
      "Epoch 7/30\n",
      "15290/15290 [==============================] - 7s 440us/step - loss: 1438.3820 - acc: 0.5950 - val_loss: 1165.0194 - val_acc: 0.6181\n",
      "Epoch 8/30\n",
      "15290/15290 [==============================] - 8s 517us/step - loss: 993.5015 - acc: 0.6102 - val_loss: 807.3107 - val_acc: 0.6367\n",
      "Epoch 9/30\n",
      "15290/15290 [==============================] - 8s 500us/step - loss: 690.6465 - acc: 0.6229 - val_loss: 565.2326 - val_acc: 0.6393\n",
      "Epoch 10/30\n",
      "15290/15290 [==============================] - 8s 546us/step - loss: 481.9598 - acc: 0.6378 - val_loss: 396.3738 - val_acc: 0.6479\n",
      "Epoch 11/30\n",
      "15290/15290 [==============================] - 7s 433us/step - loss: 337.0333 - acc: 0.6483 - val_loss: 281.7913 - val_acc: 0.6576\n",
      "Epoch 12/30\n",
      "15290/15290 [==============================] - 7s 427us/step - loss: 237.3828 - acc: 0.6564 - val_loss: 198.4267 - val_acc: 0.6673\n",
      "Epoch 13/30\n",
      "15290/15290 [==============================] - 6s 415us/step - loss: 167.4914 - acc: 0.6579 - val_loss: 141.0932 - val_acc: 0.6736\n",
      "Epoch 14/30\n",
      "15290/15290 [==============================] - 6s 421us/step - loss: 118.6145 - acc: 0.6640 - val_loss: 101.3215 - val_acc: 0.6772\n",
      "Epoch 15/30\n",
      "15290/15290 [==============================] - 8s 494us/step - loss: 84.9440 - acc: 0.6689 - val_loss: 73.0315 - val_acc: 0.6845\n",
      "Epoch 16/30\n",
      "15290/15290 [==============================] - 6s 404us/step - loss: 60.9974 - acc: 0.6740 - val_loss: 53.2763 - val_acc: 0.6869\n",
      "Epoch 17/30\n",
      "15290/15290 [==============================] - 6s 402us/step - loss: 44.3348 - acc: 0.6788 - val_loss: 39.4095 - val_acc: 0.6877\n",
      "Epoch 18/30\n",
      "15290/15290 [==============================] - 6s 398us/step - loss: 32.8778 - acc: 0.6833 - val_loss: 29.4528 - val_acc: 0.6921\n",
      "Epoch 19/30\n",
      "15290/15290 [==============================] - 7s 476us/step - loss: 24.6849 - acc: 0.6888 - val_loss: 22.7498 - val_acc: 0.6874\n",
      "Epoch 20/30\n",
      "15290/15290 [==============================] - 8s 535us/step - loss: 18.8769 - acc: 0.7006 - val_loss: 17.5699 - val_acc: 0.7021\n",
      "Epoch 21/30\n",
      "15290/15290 [==============================] - 9s 566us/step - loss: 15.0199 - acc: 0.7112 - val_loss: 15.2910 - val_acc: 0.6981\n",
      "Epoch 22/30\n",
      "15290/15290 [==============================] - 7s 445us/step - loss: 12.4412 - acc: 0.7173 - val_loss: 11.5499 - val_acc: 0.7110\n",
      "Epoch 23/30\n",
      "15290/15290 [==============================] - 7s 432us/step - loss: 9.9004 - acc: 0.7347 - val_loss: 10.2593 - val_acc: 0.7117\n",
      "Epoch 24/30\n",
      "15290/15290 [==============================] - 8s 551us/step - loss: 8.4235 - acc: 0.7499 - val_loss: 8.4853 - val_acc: 0.7136\n",
      "Epoch 25/30\n",
      "15290/15290 [==============================] - 7s 477us/step - loss: 7.5167 - acc: 0.7657 - val_loss: 10.1789 - val_acc: 0.7099\n",
      "Epoch 26/30\n",
      "15290/15290 [==============================] - 10s 671us/step - loss: 8.2449 - acc: 0.7434 - val_loss: 7.2165 - val_acc: 0.7083\n",
      "Epoch 27/30\n",
      "15290/15290 [==============================] - 8s 507us/step - loss: 6.3055 - acc: 0.7679 - val_loss: 6.3358 - val_acc: 0.7151\n",
      "Epoch 28/30\n",
      "15290/15290 [==============================] - 8s 491us/step - loss: 6.0304 - acc: 0.7670 - val_loss: 6.8001 - val_acc: 0.6123\n",
      "Epoch 29/30\n",
      "15290/15290 [==============================] - 8s 518us/step - loss: 7.1322 - acc: 0.7508 - val_loss: 9.0569 - val_acc: 0.7107\n",
      "Epoch 30/30\n",
      "15290/15290 [==============================] - 8s 507us/step - loss: 6.1608 - acc: 0.7827 - val_loss: 7.8484 - val_acc: 0.6900\n"
     ]
    }
   ],
   "source": [
    "#without StandardScaler\n",
    "history1 = model.fit(X_train, y_train, epochs=30, batch_size=256, validation_data=(X_test,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's visualize the model accuracy in the training phase: we see that both the training and test accuracy improve till the 26th epoch, then the test accuracy starts to drop**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:24:10.322923Z",
     "start_time": "2019-10-29T11:24:09.825506Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history1.history['acc'])\n",
    "plt.plot(history1.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We reach a 69% accuracy on the test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:24:21.379838Z",
     "start_time": "2019-10-29T11:24:20.450417Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3823/3823 [==============================] - 1s 212us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.619167909677073, 0.6900340047395264]"
      ]
     },
     "execution_count": 148,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Neural Network after scaling numerical features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We now scale the numerical features and see how this modification impacts the model performance, and repeat the previous transformations**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:01:11.767500Z",
     "start_time": "2019-10-29T11:01:06.465688Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#le = LabelEncoder()\n",
    "scaler = StandardScaler()\n",
    "\n",
    "dfp = df.toPandas()\n",
    "col_y = le.fit_transform(dfp.winner)\n",
    "\n",
    "#scaling and substituting\n",
    "for col_name in [\"turns\", \"white_rating\", \"black_rating\", \"opening_ply\", \"increment1\", \"increment2\"]:\n",
    "    dfp[col_name]= scaler.fit_transform(dfp[col_name].values.reshape(-1, 1))\n",
    "\n",
    "dfp2 = dfp.drop([\"victory_status\", \"winner\", \"rated\"], 1)\n",
    "dfp3 = pd.get_dummies(dfp2)\n",
    "dfp3[\"winner\"] = col_y\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(dfp3.drop([\"winner\"],1), \n",
    "                                                    dfp3.winner, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's see the scaled dataframe**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:01:11.839435Z",
     "start_time": "2019-10-29T11:01:11.790164Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "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>rated</th>\n",
       "      <th>turns</th>\n",
       "      <th>victory_status</th>\n",
       "      <th>winner</th>\n",
       "      <th>white_rating</th>\n",
       "      <th>black_rating</th>\n",
       "      <th>opening_name</th>\n",
       "      <th>opening_ply</th>\n",
       "      <th>increment1</th>\n",
       "      <th>increment2</th>\n",
       "      <th>moves_new0</th>\n",
       "      <th>moves_new1</th>\n",
       "      <th>moves_new2</th>\n",
       "      <th>moves_new3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>True</td>\n",
       "      <td>2.343753</td>\n",
       "      <td>draw</td>\n",
       "      <td>draw</td>\n",
       "      <td>0.478251</td>\n",
       "      <td>1.043109</td>\n",
       "      <td>Old Benoni Defense</td>\n",
       "      <td>-1.006279</td>\n",
       "      <td>-0.221724</td>\n",
       "      <td>-0.372712</td>\n",
       "      <td>d4</td>\n",
       "      <td>c5</td>\n",
       "      <td>Rh3</td>\n",
       "      <td>Kd2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>True</td>\n",
       "      <td>1.059687</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>2.109200</td>\n",
       "      <td>1.563021</td>\n",
       "      <td>Sicilian Defense: Alapin Variation</td>\n",
       "      <td>-0.648908</td>\n",
       "      <td>-0.338875</td>\n",
       "      <td>-0.372712</td>\n",
       "      <td>e4</td>\n",
       "      <td>c5</td>\n",
       "      <td>Re2</td>\n",
       "      <td>a3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>True</td>\n",
       "      <td>-0.075068</td>\n",
       "      <td>resign</td>\n",
       "      <td>black</td>\n",
       "      <td>0.188611</td>\n",
       "      <td>0.116909</td>\n",
       "      <td>Scotch Game: Malaniuk Variation</td>\n",
       "      <td>1.137949</td>\n",
       "      <td>-0.456026</td>\n",
       "      <td>0.206652</td>\n",
       "      <td>e4</td>\n",
       "      <td>e5</td>\n",
       "      <td>h3</td>\n",
       "      <td>Bd5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>False</td>\n",
       "      <td>-0.642446</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>0.740306</td>\n",
       "      <td>0.791761</td>\n",
       "      <td>Queen's Pawn Game: London System</td>\n",
       "      <td>0.065835</td>\n",
       "      <td>-0.221724</td>\n",
       "      <td>-0.372712</td>\n",
       "      <td>d4</td>\n",
       "      <td>d5</td>\n",
       "      <td>h6</td>\n",
       "      <td>Qc4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>True</td>\n",
       "      <td>0.970101</td>\n",
       "      <td>resign</td>\n",
       "      <td>white</td>\n",
       "      <td>2.136785</td>\n",
       "      <td>1.308230</td>\n",
       "      <td>King's Pawn Game: Nimzowitsch Defense</td>\n",
       "      <td>-0.648908</td>\n",
       "      <td>-0.221724</td>\n",
       "      <td>-0.372712</td>\n",
       "      <td>e4</td>\n",
       "      <td>Nc6</td>\n",
       "      <td>Rd2+</td>\n",
       "      <td>Bxd2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   rated     turns victory_status winner  white_rating  black_rating  \\\n",
       "0   True  2.343753           draw   draw      0.478251      1.043109   \n",
       "1   True  1.059687         resign  black      2.109200      1.563021   \n",
       "2   True -0.075068         resign  black      0.188611      0.116909   \n",
       "3  False -0.642446         resign  white      0.740306      0.791761   \n",
       "4   True  0.970101         resign  white      2.136785      1.308230   \n",
       "\n",
       "                            opening_name  opening_ply  increment1  increment2  \\\n",
       "0                     Old Benoni Defense    -1.006279   -0.221724   -0.372712   \n",
       "1     Sicilian Defense: Alapin Variation    -0.648908   -0.338875   -0.372712   \n",
       "2        Scotch Game: Malaniuk Variation     1.137949   -0.456026    0.206652   \n",
       "3       Queen's Pawn Game: London System     0.065835   -0.221724   -0.372712   \n",
       "4  King's Pawn Game: Nimzowitsch Defense    -0.648908   -0.221724   -0.372712   \n",
       "\n",
       "  moves_new0 moves_new1 moves_new2 moves_new3  \n",
       "0         d4         c5        Rh3        Kd2  \n",
       "1         e4         c5        Re2         a3  \n",
       "2         e4         e5         h3        Bd5  \n",
       "3         d4         d5         h6        Qc4  \n",
       "4         e4        Nc6       Rd2+       Bxd2  "
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dfp.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's train it, this time for 20 epochs, with a batch size of 128**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:13:09.900131Z",
     "start_time": "2019-10-29T11:09:20.929388Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15290 samples, validate on 3823 samples\n",
      "Epoch 1/20\n",
      "15290/15290 [==============================] - 13s 863us/step - loss: 1.5176 - acc: 0.5857 - val_loss: 0.9866 - val_acc: 0.7146\n",
      "Epoch 2/20\n",
      "15290/15290 [==============================] - 10s 673us/step - loss: 0.9287 - acc: 0.7600 - val_loss: 1.0151 - val_acc: 0.7264\n",
      "Epoch 3/20\n",
      "15290/15290 [==============================] - 11s 691us/step - loss: 0.8530 - acc: 0.8101 - val_loss: 1.1182 - val_acc: 0.7154\n",
      "Epoch 4/20\n",
      "15290/15290 [==============================] - 12s 755us/step - loss: 0.8160 - acc: 0.8547 - val_loss: 1.2339 - val_acc: 0.7086\n",
      "Epoch 5/20\n",
      "15290/15290 [==============================] - 13s 858us/step - loss: 0.7902 - acc: 0.8776 - val_loss: 1.3570 - val_acc: 0.7086\n",
      "Epoch 6/20\n",
      "15290/15290 [==============================] - 12s 801us/step - loss: 0.7404 - acc: 0.9002 - val_loss: 1.3812 - val_acc: 0.6830\n",
      "Epoch 7/20\n",
      "15290/15290 [==============================] - 10s 685us/step - loss: 0.7279 - acc: 0.9122 - val_loss: 1.4471 - val_acc: 0.6945\n",
      "Epoch 8/20\n",
      "15290/15290 [==============================] - 9s 620us/step - loss: 0.6892 - acc: 0.9264 - val_loss: 1.5650 - val_acc: 0.6715\n",
      "Epoch 9/20\n",
      "15290/15290 [==============================] - 10s 669us/step - loss: 0.6860 - acc: 0.9280 - val_loss: 1.5555 - val_acc: 0.6906\n",
      "Epoch 10/20\n",
      "15290/15290 [==============================] - 12s 762us/step - loss: 0.6593 - acc: 0.9337 - val_loss: 1.5420 - val_acc: 0.6754\n",
      "Epoch 11/20\n",
      "15290/15290 [==============================] - 11s 731us/step - loss: 0.6345 - acc: 0.9393 - val_loss: 1.6545 - val_acc: 0.6882\n",
      "Epoch 12/20\n",
      "15290/15290 [==============================] - 11s 726us/step - loss: 0.6424 - acc: 0.9394 - val_loss: 1.6213 - val_acc: 0.6798\n",
      "Epoch 13/20\n",
      "15290/15290 [==============================] - 11s 732us/step - loss: 0.6309 - acc: 0.9441 - val_loss: 1.5878 - val_acc: 0.6950\n",
      "Epoch 14/20\n",
      "15290/15290 [==============================] - 11s 698us/step - loss: 0.6067 - acc: 0.9478 - val_loss: 1.6762 - val_acc: 0.6864\n",
      "Epoch 15/20\n",
      "15290/15290 [==============================] - 10s 660us/step - loss: 0.6041 - acc: 0.9508 - val_loss: 1.6451 - val_acc: 0.6963\n",
      "Epoch 16/20\n",
      "15290/15290 [==============================] - 11s 723us/step - loss: 0.5916 - acc: 0.9504 - val_loss: 1.6099 - val_acc: 0.6882\n",
      "Epoch 17/20\n",
      "15290/15290 [==============================] - 11s 724us/step - loss: 0.5828 - acc: 0.9518 - val_loss: 1.6314 - val_acc: 0.6895\n",
      "Epoch 18/20\n",
      "15290/15290 [==============================] - 12s 808us/step - loss: 0.5688 - acc: 0.9551 - val_loss: 1.7095 - val_acc: 0.6840\n",
      "Epoch 19/20\n",
      "15290/15290 [==============================] - 11s 727us/step - loss: 0.5743 - acc: 0.9510 - val_loss: 1.6933 - val_acc: 0.6722\n",
      "Epoch 20/20\n",
      "15290/15290 [==============================] - 13s 858us/step - loss: 0.5436 - acc: 0.9572 - val_loss: 1.8051 - val_acc: 0.6730\n"
     ]
    }
   ],
   "source": [
    "#with StandardScaler\n",
    "history2 = model.fit(X_train, y_train, epochs=20, batch_size=128, validation_data=(X_test,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Here we see that the training accuracy increases steadily, while the test accuracy starts at a high level and slightly drops during the subsequent epochs, so less epochs are required. We reach 72% of accuracy on the test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:13:13.047883Z",
     "start_time": "2019-10-29T11:13:12.320574Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history2.history['acc'])\n",
    "plt.plot(history2.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**Let's repeat the process here, reducing the number of epochs to just 5**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:15:17.212203Z",
     "start_time": "2019-10-29T11:14:13.036469Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15290 samples, validate on 3823 samples\n",
      "Epoch 1/5\n",
      "15290/15290 [==============================] - 16s 1ms/step - loss: 1.4989 - acc: 0.5768 - val_loss: 1.0055 - val_acc: 0.7031\n",
      "Epoch 2/5\n",
      "15290/15290 [==============================] - 12s 782us/step - loss: 0.9270 - acc: 0.7542 - val_loss: 1.0049 - val_acc: 0.7185\n",
      "Epoch 3/5\n",
      "15290/15290 [==============================] - 11s 725us/step - loss: 0.8554 - acc: 0.8076 - val_loss: 1.1710 - val_acc: 0.6921\n",
      "Epoch 4/5\n",
      "15290/15290 [==============================] - 11s 699us/step - loss: 0.8206 - acc: 0.8449 - val_loss: 1.1969 - val_acc: 0.7063\n",
      "Epoch 5/5\n",
      "15290/15290 [==============================] - 11s 732us/step - loss: 0.7833 - acc: 0.8772 - val_loss: 1.3178 - val_acc: 0.7002\n"
     ]
    }
   ],
   "source": [
    "#with StandardScaler, less epochs\n",
    "history3 = model.fit(X_train, y_train, epochs=5, batch_size=128, validation_data=(X_test,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:15:19.850964Z",
     "start_time": "2019-10-29T11:15:19.319946Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nO3deXxU9bn48c+TjQRICCRhTUJYEvY9AqKCiguutJUq+vPnUi21v+ty67VKW6tWbau+ulnltqLSautye7VaahXUquDGLgJhJ2wBAiEsCSH7PL8/zglMwgADZOZMMs/79ZpXzpxl5slJ5jzzXY+oKsYYY0xTMV4HYIwxJjJZgjDGGBOQJQhjjDEBWYIwxhgTkCUIY4wxAVmCMMYYE5AlCBP1RCRHRFRE4oLY9xYR+SwccRnjNUsQpkURkS0iUiMi6U3Wf+Ve5HO8icyY1scShGmJNgPXNzwRkSFAW+/CiQzBlICMORWWIExL9BfgJr/nNwMv++8gIh1E5GURKRGRrSLyoIjEuNtiReRXIrJXRAqBKwIc+6KI7BKRHSLyuIjEBhOYiPyviBSLyEERmS8ig/y2JYnIr914DorIZyKS5G47V0S+EJEDIrJdRG5x138iIrf7vUajKi631PQfIrIB2OCue9p9jTIRWSoi5/ntHysiPxaRTSJS7m7PEpEZIvLrJr/LbBH5QTC/t2mdLEGYlmgBkCIiA9wL91Tgr032eQboAPQGJuAklFvdbd8FrgRGAPnAlCbH/hmoA/q6+1wC3E5w3gNygc7AMuAVv22/AkYB44BOwP2AT0R6usc9A2QAw4HlQb4fwDeAMcBA9/li9zU6Aa8C/ysiie62e3FKX5cDKcB3gMPAS8D1fkk0HbjIPd5EK1W1hz1azAPYgnPhehD4JTAJ+ACIAxTIAWKBGmCg33HfAz5xlz8C7vDbdol7bBzQBagGkvy2Xw987C7fAnwWZKyp7ut2wPkyVgkMC7Dfj4C3jvManwC3+z1v9P7u6194kjj2N7wvsA6YfJz91gAXu8t3Au96/fe2h7cPq7M0LdVfgPlAL5pULwHpQDyw1W/dVqCHu9wd2N5kW4Oe7rG7RKRhXUyT/QNySzM/B76NUxLw+cXTBkgENgU4NOs464PVKDYRuQ+4Def3VJySQkOj/one6yXgRpyEeyPw9BnEZFoBq2IyLZKqbsVprL4c+HuTzXuBWpyLfYNsYIe7vAvnQum/rcF2nBJEuqqmuo8UVR3Eyd0ATMYp4XTAKc0AiBtTFdAnwHHbj7MeoILGDfBdA+xzZEpmt73hfuBaoKOqpgIH3RhO9l5/BSaLyDBgAPD2cfYzUcIShGnJbsOpXqnwX6mq9cDfgJ+LSLJbx38vR9sp/gbcLSKZItIRmO537C7gfeDXIpIiIjEi0kdEJgQRTzJOcinFuaj/wu91fcAs4Dci0t1tLD5bRNrgtFNcJCLXikiciKSJyHD30OXAt0SkrYj0dX/nk8VQB5QAcSLyEE4JosELwGMikiuOoSKS5sZYhNN+8RfgTVWtDOJ3Nq2YJQjTYqnqJlVdcpzNd+F8+y4EPsNpbJ3lbnsemAt8jdOQ3LQEchOQAKzGqb9/A+gWREgv41RX7XCPXdBk+33ASpyL8D7gSSBGVbfhlIT+y12/HBjmHvNbnPaU3ThVQK9wYnOBOcB6N5YqGldB/QYnQb4PlAEvAkl+218ChuAkCRPlRNVuGGSMcYjIeJySVk+1i0PUsxKEMQYAEYkH7gFesORgwBKEMQYQkQHAAZyqtN95HI6JEFbFZIwxJiArQRhjjAmo1QyUS09P15ycHK/DMMaYFmXp0qV7VTUj0LZWkyBycnJYsuR4PR6NMcYEIiJbj7fNqpiMMcYEZAnCGGNMQJYgjDHGBNRq2iACqa2tpaioiKqqKq9DCZvExEQyMzOJj4/3OhRjTAvXqhNEUVERycnJ5OTk4Dd1c6ulqpSWllJUVESvXr28DscY08K16iqmqqoq0tLSoiI5AIgIaWlpUVViMsaETqtOEEDUJIcG0fb7GmNCp1VXMRljTGu280Al89eX4FO4YUz2yQ84RZYgQqi0tJSJEycCUFxcTGxsLBkZzoDFRYsWkZCQcNLXuPXWW5k+fTr9+vULaazGmMhXVVvPos37mL++hHnrS9iw5xAAI7JTLUG0NGlpaSxfvhyARx55hPbt23Pfffc12qfh5uAxMYFr+/70pz+FPE5jTGRSVTaVVBxJCAs3l1JV6yMhNobRvTpxbX4W4/MyyOvSPiTvbwnCAxs3buTqq69mxIgRfPXVV3zwwQf87Gc/Y9myZVRWVnLdddfx0EMPAXDuuefy7LPPMnjwYNLT07njjjt47733aNu2Lf/4xz/o3Lmzx7+NMaY5lVfV8vnGUuZvKGHeuhJ2HHDu/No7vR1Tz8pmQl4GY3p3om1C6C/fUZMgfvbPAlbvLGvW1xzYPYWHrwrmXvbHWrt2LS+//DL5+fkAPPHEE3Tq1Im6ujouuOACpkyZwsCBAxsdc/DgQSZMmMATTzzBvffey6xZs5g+fXqglzfGtBA+n1Kws+xIQli2bT91PqVdQizj+qbz/fP7MCEvg6xObcMeW9QkiEjTp0+fI8kB4LXXXuPFF1+krq6OnTt3snr16mMSRFJSEpdddhkAo0aN4tNPPw1rzMaY5rH3UDWfugnh0w17Ka2oAWBQ9xSmje/N+LwMRmZ3JCHO246mUZMgTvebfqi0a9fuyPKGDRt4+umnWbRoEampqdx4440BxzL4N2rHxsZSV1cXlliNMWemtt7Hsq37mbe+hPkbSli1w6nN6NQugfG56YzPy+C83Awyktt4HGljUZMgIllZWRnJycmkpKSwa9cu5s6dy6RJk7wOyxhzBrbvO+wkhPUlfLGplEPVdcTGCKOyO3LfJXlMyOvMoO4pxMRE7tglSxARYOTIkQwcOJD+/fvTs2dPzjnnHK9DMsacosqaehZsLmXeOqeUUFhSAUCP1CSuGtadCXkZjOubRkpiy5knrdXckzo/P1+b3jBozZo1DBgwwKOIvBOtv7cx4aSqbNhz6EhCWLh5HzV1PtrExTC2dxoT8jIYn5dBn4x2ET3DgYgsVdX8QNusBGGMMUE6eLiWzzbuZb7blrDroNNWmNu5PTeN7cn4vAxG9+pEYnysx5E2D0sQxhhzHPU+ZUXRAeav38u89XtYvv0APoXkxDjO7ZvOPROdUkL31CSvQw0JSxDGGONnT1kV89yRy59t3MuBw7WIwNAeHbjzgr6Mz8tgeFYqcbGtfq5TSxDGmOhWXVfP0i37meeOS1hbXA5ARnIbJvbvwvi8dM7LzaBTu5PPndbaWIIwxkSdLXsrjoxc/rKwlMM19cTHCvk9O/HApP5MyMtgQLfkiG5cDgdLEMaYVq+iuo4vN5UeGai2tfQwANmd2nLNyEwm5GUwtk8a7dvYJdFfSM+GiEwCngZigRdU9Ykm27OBl4BUd5/pqvquiOQAa4B17q4LVPWOUMYaCs0x3TfArFmzuPzyy+natWvIYjWmNVFV1uwqPzJQbcnWfdTWK20TYjm7dxrfOacXE/IyyElvd/IXi2IhSxAiEgvMAC4GioDFIjJbVVf77fYg8DdV/YOIDATeBXLcbZtUdXio4guHYKb7DsasWbMYOXKkJQhjTmB/RQ2fbtx7ZFxCSXk1AP27Jh9JCKNyOtImrnV0QQ2HUJYgRgMbVbUQQEReByYD/glCgRR3uQOwM4TxRJSXXnqJGTNmUFNTw7hx43j22Wfx+XzceuutLF++HFVl2rRpdOnSheXLl3PdddeRlJR0SiUPY1qzunofy7cfcO6VsGEvK4oOoAqpbeM5t2/6kYFqXVISvQ61xQplgugBbPd7XgSMabLPI8D7InIX0A64yG9bLxH5CigDHlTVY6YuFZFpwDSA7OyT3E3pvelQvPLUfoOT6ToELnvi5Ps1sWrVKt566y2++OIL4uLimDZtGq+//jp9+vRh7969rFzpxHngwAFSU1N55plnePbZZxk+vEUXqIw5Yw232Jy/wZkFtbyqjhiB4Vmp/OfEPMbnpTM0M5XYCJ7fqCXxukXmeuDPqvprETkb+IuIDAZ2AdmqWioio4C3RWSQqja6oYOqzgRmgjPVRriDP10ffvghixcvPjLdd2VlJVlZWVx66aWsW7eOu+++myuuuIJLLrnE40iN8dbxbrHZNSWRywd3Y3xeBuf2TadD25Yzv1FLEsoEsQPI8nue6a7zdxswCUBVvxSRRCBdVfcA1e76pSKyCcgDlnC6TuObfqioKt/5znd47LHHjtm2YsUK3nvvPWbMmMGbb77JzJkzPYjQGO9s3lvBx2v3HHOLzTG9nVtsTuiXQW7n9lHfBTUcQpkgFgO5ItILJzFMBW5oss82YCLwZxEZACQCJSKSAexT1XoR6Q3kAoUhjDWsLrroIqZMmcI999xDeno6paWlVFRUkJSURGJiIt/+9rfJzc3l9ttvByA5OZny8nKPozYmNFSVtcXlzFlVzJxVxazb7fyve3GLTdNYyM64qtaJyJ3AXJwurLNUtUBEHgWWqOps4L+A50XkBzgN1reoqorIeOBREakFfMAdqrovVLGG25AhQ3j44Ye56KKL8Pl8xMfH88c//pHY2Fhuu+02VBUR4cknnwTg1ltv5fbbb7dGatNqqCpfFx3kvVW7mLuqmC2lhxGBs3I68dCVA7l4YBdPbrFpGrPpvluhaP29TWSr9ylLtuzjvVXFzC0oZtfBKuJihLP7pHHZ4G5cPLBLxN1RLRrYdN/GGE/U1Pn4srCUOauK+WB1MXsP1ZAQF8OEvAzuu6QfFw3oYg3MEcwShDGmWVXV1jN/fQlzVhXz4ZrdlFXV0S4hlgv6d2bS4K5c0K8z7WxKixah1f+VGurzo0VrqTI0Lcuh6jo+WruHuauK+XjdHg7X1NMhKZ6LB3blssFdOTc3vdXcRCeatOoEkZiYSGlpKWlpaVGRJFSV0tJSEhNt5KgJvf0VNXy4ZjdzVhXz6ca91NT5SG/fhm+O6MGkwV0Z2zuN+Ci4Z0Jr1qoTRGZmJkVFRZSUlHgdStgkJiaSmZnpdRimldpTXsXcgt3MXVXMl4Wl1PuUHqlJ3DimJ5cN6crI7I42irkVadUJIj4+nl69enkdhjEtWtH+w0fGKCzdth9V6JXejmnje3PZ4K4M6dEhKkro0ahVJwhjzOnZVHLoSFJYueMg4MyK+p8T85g0uCt5XWwkczSwBGGMOXL/hDmrdjGnoJj1u505j4ZnpTL9sv5MGtTV7p0QhSxBGBOlfD5ledEB5q4qZk5BMVtLDxPjjmZ+5KqBXDKoK91Tk7wO03jIEoQxUaTepyzavI85q3Yxt2A3xWVVxMcK4/qkc8eEPlw8sAvp7W00s3FYgjCmlaup8/HFpr3uaObdlFbU0MYdzfzAkH5c2L8LHZJsNLM5liUIY1qhypp65q0vYW6BM5q5vKqO9m3iuNAdzXx+vwybHdWclP2HGNNKlFfV8tHaPcxZVcwn60qorK0ntW08kwZ1ZdLgrpzT10Yzm1NjCcKYFmx/RQ0frN7NnIJiPtuwl5p6HxnJbbhmVA8uG9yN0b062Whmc9osQRjTwuwuq+L9gmLeW1XMws37joxmvunsnkwa7IxmjrHRzKYZWIIwpgXYvs8dzVxQzDJ3NHPvjHbcMaE3lw3uxqDuKTZwzTQ7SxDGRKiNew4xZ9Uu3ltVTMHOMgAGdkvhBxflcdngruR2SfY4QtPaWYIwJkKoKgU7y5jrVh9t3OOMZh6RncqPL+/PpYO60jPNRjOb8LEEYYyHfD7lq+0HjkxxsX1fJTECY3ql8X/H9uTSQV3p2sGmbzfesARhTJjV1fuc0cwFzr2Zd5dVEx8rnNM3nTsv6MtFA7qQZqOZTQSwBGFMGFTX1fPFRvfezGt2s6+ihsR4ZzTzZYO7ceGAzqQk2mhmE1ksQRgTQmuLy3jh083MXVVMebUzmnnigM5MGtSVCTaa2UQ4++80ppmpKgs37+O5eZv4eF0JbRNiuXJoNy4b3I1xfdNoE2ejmU3LYAnCmGZS71M+WF3MH+YV8vX2A6S1S+C+S/K4cWxPUtsmeB2eMacspAlCRCYBTwOxwAuq+kST7dnAS0Cqu890VX3X3fYj4DagHrhbVeeGMlZjTldVbT1/X7aD5z8tZPPeCnqmteXxbwxmyqhMm/vItGghSxAiEgvMAC4GioDFIjJbVVf77fYg8DdV/YOIDATeBXLc5anAIKA78KGI5KlqfajiNeZUHTxcy18XbuVPn29h76FqhvTowIwbRjJpcFdibaoL0wqEsgQxGtioqoUAIvI6MBnwTxAKpLjLHYCd7vJk4HVVrQY2i8hG9/W+DGG8xgRl18FKXvx0M68t2kZFTT3j8zK4Y0Jvzu6dZtNdmFYllAmiB7Dd73kRMKbJPo8A74vIXUA74CK/Yxc0ObZH0zcQkWnANIDs7OxmCdqY41m/u5zn5hXyj+U7UODKod343vg+DOyectJjjWmJvG6kvh74s6r+WkTOBv4iIoODPVhVZwIzAfLz8zVEMZoopqos3rKf5+Zt4t9r95AUH8uNY3ty27m9yOrU1uvwjAmpUCaIHUCW3/NMd52/24BJAKr6pYgkAulBHmtMyPh8yvurd/Pc/E18te0Andol8IOL8rjp7J50bGc9kkx0CGWCWAzkikgvnIv7VOCGJvtsAyYCfxaRAUAiUALMBl4Vkd/gNFLnAotCGKsxgDPi+a1lO5g5v5DCvRVkdUriscmDmDIqi6QE65FkokvIEoSq1onIncBcnC6ss1S1QEQeBZao6mzgv4DnReQHOA3Wt6iqAgUi8jecBu064D+sB5MJpYOVtby6cBuzPt9MSXk1g7qn8Mz1I7hscFfi7I5sJkqJcz1u+fLz83XJkiVeh2FamOKDVcz6fDOvLtzGoeo6zstN53vj+3BOX+uRZKKDiCxV1fxA27xupDbGExt2lzNzfiFvL99BvU+5cmh3po3vzeAeHbwOzZiIYQnCRJUlW/bxx3mb+HDNHhLjY7hhdDa3n9fbeiQZE4AlCNPq+XzKh2t289z8QpZu3U/HtvHcMzGXm8fl0Ml6JBlzXJYgTKtVXVfPP77ayXPzN7GppILMjkn87OpBfDs/06bZNiYI9ikxrU5Zldsj6bPN7CmvZmC3FJ6eOpwrhnSzHknGnAJLEKbV2F3m9khasI3y6jrO6ZvGr68dxrl9061HkjGnwRKEafE27jnE8/MLeeurHdT5fFw+xJkjaUim9Ugy5kxYgjAt1tKt+/njvE18sHo3beJiuO6sLG4/rxc909p5HZoxrYIlCNOi+HzKR2v38Nz8TSzesp/UtvHcPTGXm8/uSVr7Nl6HZ0yrYgnCtAg1dT7+sdyZI2nDnkP0SE3i4asGcm1+Fu3a2L+xMaFgnywT0cqranlt0TZmfbaF4rIq+ndN5nfXDeeKod2Itx5JxoSUJQgTkfaUVfGnL7bw1wVbKa+q4+zeaTw5ZSjjc61HkjHhYgnCRJRNJU6PpL8vc3okXTa4G9PG92ZYVqrXoRkTdSxBmIiwbJtz17b3V+8mPjaGb+dn8t3zepOTbj2SjPGKJQjjGZ9P+WT9Hv44r5BFm/fRISmeOy/oy01n55CRbD2SjPGaJQgTdjV1Pv75tTNH0vrdh+jeIZGfXjmQqWdZjyRjIol9Gk3YHKqu4/VF23jxs83sOlhFvy7J/ObaYVw1rLv1SDImAlmCMCFXUl7Nn7/YzF++3EpZVR1je3fiF98awvl5GdYjyZgIZgnChMzmvRXMnF/Im8uKqK33MWlQV6aN782I7I5eh2aMCYIlCNPslm8/wHPzNjGnoJj42BiuGZnJtPG96WU9koxpUSxBmGahqnyyvoTn5m1iQeE+khPj+P6EPtxyTg6dkxO9Ds8YcxosQZgzUlvv9EiaOb+QtcXldE1J5MErBjB1dDbtrUeSMS3aST/BInIX8FdV3R+GeEwLUVFdx+uLt/Pip4XsPFhFbuf2/Orbw7h6WHcS4qxHkjGtQTBf8boAi0VkGTALmKuqGtqwTKRSVf4wbxPPzSvkYGUto3M68fg3B3N+XmdiYqxHkjGtyUkThKo+KCI/BS4BbgWeFZG/AS+q6qYTHSsik4CngVjgBVV9osn23wIXuE/bAp1VNdXdVg+sdLdtU9Wrg/+1TKi8s2IXT81ZxwX9MrjzwlxG9bQeSca0VkFVEquqikgxUAzUAR2BN0TkA1W9P9AxIhILzAAuBopwSiGzVXW13+v+wG//u4ARfi9RqarDT/UXMqGzr6KGR2YXMDSzA8/flE+cDW4zplU76SdcRO4RkaXAU8DnwBBV/T4wCrjmBIeOBjaqaqGq1gCvA5NPsP/1wGtBR27C7tF/FnCwspanpgy15GBMFAimBNEJ+JaqbvVfqao+EbnyBMf1ALb7PS8CxgTaUUR6Ar2Aj/xWJ4rIEpwSyxOq+naA46YB0wCys7OD+FXM6fpo7W7eXr6Tuyfm0r9ritfhGGPCIJivge8B+xqeiEiKiIwBUNU1zRTHVOANVa33W9dTVfOBG4DfiUifpgep6kxVzVfV/IyMjGYKxTRVVlXLj/++irwu7bnzgr5eh2OMCZNgEsQfgEN+zw+5605mB5Dl9zzTXRfIVJpUL6nqDvdnIfAJjdsnTBj98t217Cmv4qkpw6wLqzFRJJhPu/h3a1VVH8FVTS0GckWkl4gk4CSB2ce8uEh/nEbvL/3WdRSRNu5yOnAOsLrpsSb0vti0l9cWbeO2c3sx3O7qZkxUCSZBFIrI3SIS7z7uAQpPdpCq1gF3AnOBNcDfVLVARB4VEf8uq1OB15uMrRgALBGRr4GPcdogLEGE2eGaOqa/uZKctLbce3E/r8MxxoRZMCWBO4DfAw8CCvwbt2H4ZFT1XeDdJuseavL8kQDHfQEMCeY9TOj8+v31bNt3mNenjSUpIdbrcIwxYRbMQLk9ON/yTRRZtm0/sz7fzP8Zk83Y3mleh2OM8UAwczElArcBg4Aj03Kq6ndCGJfxUHVdPQ+8sYJuKYlMv6y/1+EYYzwSTBvEX4CuwKXAPJzeSOWhDMp4a8ZHG9mw5xA//9YQkhPjvQ7HGOORYBJEX1X9KVChqi8BV3CcAW+m5Vu9s4z//mQT3xrRgwv6dfY6HGOMh4JJELXuzwMiMhjoANiVoxWqq/fxwJsrSG0bz0+vHOh1OMYYjwXTi2mmiHTE6cU0G2gP/DSkURlPPP/pZlbuOMiMG0bSsV2C1+EYYzx2wgQhIjFAmXuzoPlA77BEZcJuU8khfvvhei4d1IXLh3T1OhxjTAQ4YRWTO2o64HTepvXw+ZTpb64gMS6GxyYPRsRu/GOMCa4N4kMRuU9EskSkU8Mj5JGZsPnLgq0s3rKfn145kM4piSc/wBgTFYJpg7jO/fkffusUq25qFbbvO8yTc9YyPi+DKaMyvQ7HGBNBghlJ3SscgZjwU1V+/NZKBPjFN61qyRjTWDAjqW8KtF5VX27+cEw4vbG0iE837OXRyYPI7NjW63CMMREmmCqms/yWE4GJwDLAEkQLtqesisfeWc1ZOR25cUxPr8MxxkSgYKqY7vJ/LiKpOPeXNi2UqvLTf6yiqs7HE9cMJSbmFKqWVOFwKSR2gFibhsOY1iyYEkRTFTj3jzYt1Lsri5lbsJsHJvWnT0b7Y3eorYQD22D/Fuexb/PR5f1boK4S4ttCj1GQNcZ9nAVJHcP6exhjQiuYNoh/4vRaAqdb7EDgb6EMyoTO/ooaHv7HSs7rVs93c/bA10v8Lv5uIijf1fig+HbQMQc69Ya+E6FDprPftgXw2W+h4VbiGf2PJozssc7+1vBtTIsVTAniV37LdcBWVS0KUTymudRWwYGtjb/5799C5ZY1zK/bSdv91fDnhp0FUro7SaDPhc7Pjr3cnznQLv34F/qaCtixFLYvhG0LYfXbsOwlZ1vbdDdhjHYSRrfhEG/jLIxpKYJJENuAXapaBSAiSSKSo6pbQhqZOTFVqCgJXAW0fwuU72y8f3w7DrXNZNXhTqRljmXU8JFuqaAXdMg6/Qt3QjvoNd55APh8sHfd0YSxfSGs+5ezLTbBSRINCSNrDLS3eR+NiVTS+FbQAXYQWQKMU9Ua93kC8LmqnnXCA8MsPz9flyxZ4nUYzau2qnFbwP4miaD2cOP9k7s7F/yGb/5+j/LYVC753ae0bxPHO3efS5u4MN5C9FAJFC1yqqS2L4KdX0F9tbOtYy+3SsqtmsoYADHBDPA3xjQHEVmqqvmBtgVTgohrSA4AqlrjJglzpvxLAf6PhhLBMaWAtkcv+r0vaJwEUrNPWAp44q2V7C6r4r+/Py68yQGgfQb0v8J5ANRVw66v3YSxEDb9G1a4HePadIDMfLeEMRp65EObAA3pxpiQCyZBlIjI1ao6G0BEJgN7QxtWK3JMKWBL49JAoFJAxxzoff7Ri39DqaBdxmk1+n65qZRXFm7j9nN7MSI7AnoaxbVxLv5Zo53nqs75aKiS2r4QPv4FoCAx0GXw0SqprDFOI7k1fhsTcsFUMfUBXgG6u6uKgJtUdWOIYzslnlUxqULF3mOrfxoeZTs52gmMxqWApo/UbIhPatbwKmvqmfT0fADm3DOepIQwlx5OV+UBKFriJowFULQUaiucbcndj1ZJZY2BrkNsTEZr4vMBCjEt5H+1hTujKiZV3QSMFZH27vNDzRxf5KurPvG4gIYLV4Pkbs4Fv9eEY5NA+85h/fb7mw/WsbX0MK9+d0zLSQ4ASamQe5HzAKivg92rnDaM7W5bRsFbzrYjYzJGQ9ZYp4qqrU04HLHqa6FsBxzY7nyuDm53lg9uc5/vAF+t83dtkwwJ7Z1qxjYpfssN65Ob7JMMCcnH7mNfIE5LMOMgfgE8paoH3Ocdgf9S1QdDHVzYHCkFbAncINy0FBCXdPSC32t844bhEJQCTtfy7Qd48bPNXD86m3F90r0O58zExkH34c5jzDRn3cEdR6ukti+Ez353dExGej+/UsZYSOtj1VLhUlsJB4uci32jBOAmhPJdoL7GxyR3c3rT9RgFA78BcYlQcwiqy6D6kLt8yHndmnJnubr8aGeHkwdzpwYAABVbSURBVIlLPHkSCSbRtEl2qkijRDBVTF+p6ogm65ap6siTvrjIJOBpIBZ4QVWfaLL9t8AF7tO2QGdVTXW33Yxzm1OAx1X1pRO912lXMZXthGfyj18KaPTo5Ukp4HRU19Vz1TOfUVZZx/v3jiclMQq+QdVUwI5lR0sY2xdC1UFnW9u0o2MyssZC9xE2JuN0VZU1uehv9VveDhV7Gu8vsdChB3TIhtQsJxGk+i13yDz9i259rZMoqsuPJpHq8sZJpKbJzyMJxz/5lB/bHng8MfFO4khwE0ab9n4JJND6lCb7+CWc+CTPryVn2ospVkTaqGq1+2JJwEn/miISC8wALsZpt1gsIrNVdXXDPqr6A7/97wJGuMudgIeBfJyv7kvdY/cHEe+padcZRt50bFtAQsue3XTGx5tYv/sQs27Jj47kAO6YjPOcB7hjMtY3LmWse9fZFhPvlEay/Noykrt4F3ukUIXK/cde9A9sc6uAtkPVgcbHxLZxLvKp2dBvkpsI/BJAcjenBBgKsfFOdWJzVCn66k+SRNznjRKR+7Nyv3Ou/Ndz4i/fgNMJI+hE0yS5NNonOSS9/YL5q70C/FtE/gQIcAtwwm/zrtHARlUtBBCR14HJwOrj7H89TlIAuBT4QFX3ucd+AEwCXgvifU9NbBxc9sTJ92tB1uwq478/3sg3hnfnwv5RfNGLiYHO/Z3HqJuddRV7G7djLHoevnzW2dYxp3HC6Dyg9TWU+nzON/ym1T/+y01L0wntj37rzxrjLmdBak9nuV1G6xi7EhPrTEKZ2OHMX8vnc0okxy3BlDdJROWN9ykvbpygGqpOj6f7CJj2yZnH3UQwjdRPisjXwEU4KXEuEMz80D2A7X7Pi4AxgXYUkZ44EwB+dIJjewQ4bhowDSA7OzuIkFq/unof97+xgg5J8Tx01SCvw4k87dKh/+XOA9wxGSvchLEQNn0MK/7H2dYmxWnwznLHZGTmO9/UIll9nTN+ptFF3/9nEdTXND4mqaNzoU/r60y1ciQBZDvLSR09rwZpcWJi3PaLZvhWrwp1VScuwSSmnPn7BBBsuW83TnL4NrAZeLOZ45gKvKF6sjTZmKrOBGaC0wbRzDG1SC9+tpmVOw7y7A0j6NTOxjOeVFwbZybarLOAu9wxGVuOVkltWwif/JKjYzIGuQnDHf3dISu8F8+66hM3AJftPPbbZvsuTpzdhsGAq46WBhoSQaQnvWgn4rRVxCc5g07D6LgJQkTycKp9rscZGPc/OI3aFxzvmCZ2AFl+zzPddYFMpfE9r3cA5zc59pMg3zdqFZYc4jcfrOeSgV24Ykg3r8NpmUScXmmdesGwqc66qoNQtNipktq2AJa/Coufd7Yld/ObwXYMdB16Zl0qqw8d2+3Tvy3gUHGTeGMgpYdzse85rvGFv0O20zZgjfHmNB23F5OI+IBPgdsaBsWJSKGq9g7qhUXigPU4d6DbASwGblDVgib79QfmAL3UDcZtpF4KNPSUWgaMamiTCKRVzsV0Cnw+ZerMBawtLuPDeyfQOcUuCiFTXwd7Co4mjO2LnIs5OF2gG8ZkZI+FzLOONqCqOg28jS76TUoClU3+xWMTnIu8/0XfvwE4pbv18Tdn5HR7MX0L55v9xyIyB+cuckGXpVW1TkTuxGmziAVmqWqBiDwKLGmYusN9j9fVL1Op6j4ReQwnqQA8eqLkYOCVhVtZtGUfT00Zaskh1GLjnOqabsNg9HeddWU7G89g+8Xv4bPfONvS8yAmzkkANeWNXyu+7dFv/T1G+XUBdde179I6GoBNixTMOIh2OL2PrgcuxLkX9Vuq+n7owwteNJcgivYf5tLfzmdkz468/J3RiDUoeq/mMOxc5pQwipY4VUFHxgA0tAFkO6UL+3sZD53pVBsVwKvAq+4o6m8DDwARlSCilary47dWocAvvjnEkkOkSGgLOec6D2NaqFMqu6rqflWdqaoTQxWQOTVvLtvB/PUl3H9pP7I6tezBfcaYyGKVmy3YnvIqHntnNfk9O3LT2Tleh2OMaWUsQbRgD71dQGVtPU9OGUpMjFUtGWOalyWIFuq9lbuYU1DMf16US58Mu+OaMab5WYJogQ4cruGn/yhgcI8Upp0X1LAUY4w5ZSGaYtGE0qPvrObA4Rpe/s5o4mItxxtjQsOuLi3Mx+v28PdlO/j++X0Y2D00E3QZYwxYgmhRyqtq+cnfV9K3c3vuvLCv1+EYY1o5q2JqQZ6cs5ZdZVW8ccc42sS1svsUGGMijpUgWogFhaX8dcE2bh3Xi1E9O3odjjEmCliCaAEqa+qZ/uYKsju15b5L87wOxxgTJayKqQX47Yfr2VJ6mFdvH0PbBPuTGWPCw0oQEe7r7Qd44dNCrh+dxbi+6V6HY4yJIpYgIlhNnY8H3lxB5+REfnT5AK/DMcZEGauviGD//clG1haX8+LN+aQk2l3DjDHhZSWICLWuuJwZH29k8vDuTBzQxetwjDFRyBJEBKqr93H/G1+TnBjPw1cN8jocY0yUsiqmCDTr8818XXSQ318/gk7tErwOxxgTpawEEWE2763g1++v56IBXbhqaDevwzHGRDFLEBHE51MeeHMFCXEx/Pybg+3+0sYYT1mCiCCvLNrGos37ePCKAXRJSfQ6HGNMlLMEESF2HKjkiXfXcG7fdK7Nz/I6HGOMsQQRCVSVn7y1Ep/CL781xKqWjDERIaQJQkQmicg6EdkoItOPs8+1IrJaRApE5FW/9fUistx9zA5lnF5766sdfLKuhPsn9SOrU1uvwzHGGCCE3VxFJBaYAVwMFAGLRWS2qq722ycX+BFwjqruF5HOfi9RqarDQxVfpCgpr+bRd1YzqmdHbj47x+twjDHmiFCWIEYDG1W1UFVrgNeByU32+S4wQ1X3A6jqnhDGE5Eenr2KwzX1PHnNUGJirGrJGBM5QpkgegDb/Z4Xuev85QF5IvK5iCwQkUl+2xJFZIm7/huB3kBEprn7LCkpKWne6MNgzqpdvLuymHsm5tK3c3uvwzHGmEa8HkkdB+QC5wOZwHwRGaKqB4CeqrpDRHoDH4nISlXd5H+wqs4EZgLk5+dreEM/MwcO1/Dg2wUM7JbCtPG9vQ7HGGOOEcoSxA7Av79mprvOXxEwW1VrVXUzsB4nYaCqO9yfhcAnwIgQxhp2j72zhv2Ha3hqylDiY60zmTEm8oTyyrQYyBWRXiKSAEwFmvZGehun9ICIpONUORWKSEcRaeO3/hxgNa3EJ+v28OayIu6Y0JvBPTp4HY4xxgQUsiomVa0TkTuBuUAsMEtVC0TkUWCJqs52t10iIquBeuCHqloqIuOA50TEh5PEnvDv/dSSHaqu4ydvraJPRjvuujDX63CMMea4QtoGoarvAu82WfeQ37IC97oP/32+AIaEMjavPDVnLTsPVvLGHeNIjI/1OhxjjDkuq/wOo0Wb9/Hyl1u5ZVwOo3p29DocY4w5IUsQYVJVW88Db64gq1MSP7y0n9fhGGPMSXndzTVq/PbD9WzeW8Ert4+hbYKddmNM5LMSRBisKDrA8/MLmXpWFuf0Tfc6HGOMCYoliBCrqfNx/xsryEhuw48uH+B1OMYYEzSr6wixP3yyibXF5Tx/Uz4dkuK9DscYY4JmJYgQWldczrMfb+CqYd25eGAXr8MxxphTYgkiROp9yv1vriA5MZ5HrhrodTjGGHPKLEGEyJ8+38zX2w/w8FUDSWvfxutwjDHmlFmCCIEteyv41fvruGhAZ64e1t3rcIwx5rRYgmhmPp8y/e8riI+J4fFv2P2ljTEtlyWIZvba4m0sKNzHT64YQNcOiV6HY4wxp80SRDPaeaCSX767lnF90rjurKyTH2CMMRHMEkQzUVV+8tZK6n3KE98aalVLxpgWzxJEM3l7+Q4+XlfCfZf2IzutrdfhGGPMGbME0QxKyqv52T9XMzI7lVvG5XgdjjHGNAtLEM3gkX8WcLi6nqemDCU2xqqWjDGtgyWIMzS3oJh/rdjF3RP70rdzstfhGGNMs7EEcQYOHq7lwbdXMbBbCt+b0MfrcIwxplnZbK5n4PF/rWZfRQ1/uuUs4mMt1xpjWhe7qp2m+etL+N+lRXxvfG8G9+jgdTjGGNPsLEGchorqOn7095X0yWjH3RNzvQ7HGGNCwqqYTsNTc9ay82Al//u9s0mMj/U6HGOMCYmQliBEZJKIrBORjSIy/Tj7XCsiq0WkQERe9Vt/s4hscB83hzLOU7F4yz5e+nIrN5+dQ35OJ6/DMcaYkAlZCUJEYoEZwMVAEbBYRGar6mq/fXKBHwHnqOp+Eensru8EPAzkAwosdY/dH6p4g1FVW88Db6wgs2MSP7y0n5ehGGNMyIWyBDEa2KiqhapaA7wOTG6yz3eBGQ0XflXd466/FPhAVfe52z4AJoUw1qA8/e8NFO6t4JffGkK7NlY7Z4xp3UKZIHoA2/2eF7nr/OUBeSLyuYgsEJFJp3BsWK3acZCZ8wu5Nj+T83IzvAzFGGPCwuuvwXFALnA+kAnMF5EhwR4sItOAaQDZ2dmhiA+A2nofP3xjBWntEvjJFXZ/aWNMdAhlCWIH4H9ThEx3nb8iYLaq1qrqZmA9TsII5lhUdaaq5qtqfkZG6L7V//GTTazZVcbj3xhMh6T4kL2PMcZEklAmiMVAroj0EpEEYCowu8k+b+OUHhCRdJwqp0JgLnCJiHQUkY7AJe66sNuwu5xnPtrIlUO7ccmgrl6EYIwxnghZFZOq1onInTgX9lhglqoWiMijwBJVnc3RRLAaqAd+qKqlACLyGE6SAXhUVfeFKtbjqfcpP3xjBe3axPKzqweF++2NMcZToqpex9As8vPzdcmSJc36mi98Wsjj/1rD01OHM3m4p23kxhgTEiKyVFXzA22zqTaOY2tpBb96fx0X9u/M1cO6ex2OMcaEnSWIAFSV6W+uJD4mhp9/c7DdX9oYE5UsQQTw+uLtfFlYyo8uH0C3Dkleh2OMMZ6wBNHEroOV/OJfazi7dxrXj846+QHGGNNKWYLwo6o8+NYqan0+nrhmiFUtGWOimiUIP7O/3sm/1+7hvkv60TOtndfhGGOMpyxBuPYequaR2QWMyE7l1nN6eR2OMcZ4zhKE65HZBVRU1/PUNUOJjbGqJWOMsQQBvF9QzDsrdnHXhX3J7ZLsdTjGGBMRoj5BHKys5cG3V9G/azJ3nN/H63CMMSZieD3dt+dq6nwMy0rl7gtziY+N+nxpjDFHRH2CyEhuw/M3BZyGxBhjopp9ZTbGGBOQJQhjjDEBWYIwxhgTkCUIY4wxAVmCMMYYE5AlCGOMMQFZgjDGGBOQJQhjjDEBiap6HUOzEJESYOsZvEQ6sLeZwmlOFtepsbhOjcV1alpjXD1VNSPQhlaTIM6UiCxR1YgbUm1xnRqL69RYXKcm2uKyKiZjjDEBWYIwxhgTkCWIo2Z6HcBxWFynxuI6NRbXqYmquKwNwhhjTEBWgjDGGBOQJQhjjDEBRVWCEJFJIrJORDaKyPQA29uIyP+42xeKSE6ExHWLiJSIyHL3cXuY4polIntEZNVxtouI/N6Ne4WIjIyQuM4XkYN+5+uhMMWVJSIfi8hqESkQkXsC7BP2cxZkXGE/ZyKSKCKLRORrN66fBdgn7J/JIOPy5DPpvnesiHwlIu8E2Na850tVo+IBxAKbgN5AAvA1MLDJPv8P+KO7PBX4nwiJ6xbgWQ/O2XhgJLDqONsvB94DBBgLLIyQuM4H3vHgfHUDRrrLycD6AH/LsJ+zIOMK+zlzz0F7dzkeWAiMbbKPF5/JYOLy5DPpvve9wKuB/l7Nfb6iqQQxGtioqoWqWgO8Dkxuss9k4CV3+Q1goohIBMTlCVWdD+w7wS6TgZfVsQBIFZFuERCXJ1R1l6ouc5fLgTVAjya7hf2cBRlX2Lnn4JD7NN59NO01E/bPZJBxeUJEMoErgBeOs0uznq9oShA9gO1+z4s49kNyZB9VrQMOAmkREBfANW6VxBsikhXimIIVbOxeONutInhPRAaF+83dov0InG+f/jw9ZyeICzw4Z251yXJgD/CBqh73fIXxMxlMXODNZ/J3wP2A7zjbm/V8RVOCaMn+CeSo6lDgA45+QzCBLcOZX2YY8AzwdjjfXETaA28C/6mqZeF87xM5SVyenDNVrVfV4UAmMFpEBofjfU8miLjC/pkUkSuBPaq6NNTv1SCaEsQOwD/LZ7rrAu4jInFAB6DU67hUtVRVq92nLwCjQhxTsII5p2GnqmUNVQSq+i4QLyLp4XhvEYnHuQi/oqp/D7CLJ+fsZHF5ec7c9zwAfAxMarLJi8/kSePy6DN5DnC1iGzBqYq+UET+2mSfZj1f0ZQgFgO5ItJLRBJwGnBmN9lnNnCzuzwF+Ejd1h4v42pSR301Th1yJJgN3OT2zBkLHFTVXV4HJSJdG+pdRWQ0zv95yC8q7nu+CKxR1d8cZ7ewn7Ng4vLinIlIhoikustJwMXA2ia7hf0zGUxcXnwmVfVHqpqpqjk414mPVPXGJrs16/mKO90DWxpVrRORO4G5OD2HZqlqgYg8CixR1dk4H6K/iMhGnEbQqRES190icjVQ58Z1S6jjAhCR13B6t6SLSBHwME6DHar6R+BdnF45G4HDwK0REtcU4PsiUgdUAlPDkOjB+Yb3f4GVbv01wI+BbL/YvDhnwcTlxTnrBrwkIrE4CelvqvqO15/JIOPy5DMZSCjPl021YYwxJqBoqmIyxhhzCixBGGOMCcgShDHGmIAsQRhjjAnIEoQxxpiALEEYcwpEpN5vBs/lEmD23TN47Rw5zgy1xnghasZBGNNMKt0pGIxp9awEYUwzEJEtIvKUiKx07yXQ112fIyIfuZO6/VtEst31XUTkLXdyvK9FZJz7UrEi8rw49yF43x3Ja4wnLEEYc2qSmlQxXee37aCqDgGexZl1E5yJ715yJ3V7Bfi9u/73wDx3cryRQIG7PheYoaqDgAPANSH+fYw5LhtJbcwpEJFDqto+wPotwIWqWuhOjFesqmkishfopqq17vpdqpouIiVApt+Ebw1TcX+gqrnu8weAeFV9PPS/mTHHshKEMc1Hj7N8Kqr9luuxdkLjIUsQxjSf6/x+fukuf8HRCdP+D/Cpu/xv4Ptw5OY0HcIVpDHBsm8nxpyaJL8ZUQHmqGpDV9eOIrICpxRwvbvuLuBPIvJDoISjs7feA8wUkdtwSgrfBzyfKt0Yf9YGYUwzcNsg8lV1r9exGNNcrIrJGGNMQFaCMMYYE5CVIIwxxgRkCcIYY0xAliCMMcYEZAnCGGNMQJYgjDHGBPT/AYmfY1W2qFwZAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training validation accuracy values\n",
    "plt.plot(history3.history['acc'])\n",
    "plt.plot(history3.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We now reach 70% accuracy on the test set, so the scaling did NOT improve our results**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:15:48.709940Z",
     "start_time": "2019-10-29T11:15:47.676250Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3823/3823 [==============================] - 1s 230us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.9364024672463338, 0.7002354172583871]"
      ]
     },
     "execution_count": 140,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "### Neural Network with deeper architecture"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We try a deeper architecture, with a greater number of trainable parameters and more layers**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:45:51.574265Z",
     "start_time": "2019-10-29T11:45:50.739969Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "model2 = Sequential()\n",
    "model2.add(Dense(1024, input_shape=(input_dim,), activity_regularizer=regularizers.l2(0.01)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Dropout(rate=0.2))\n",
    "model2.add(Dense(512, activity_regularizer=regularizers.l2(0.01)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Dropout(rate=0.2))\n",
    "model2.add(Dense(256))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Dropout(rate=0.2))\n",
    "model2.add(Dense(128))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Dropout(rate=0.2))\n",
    "model2.add(Dense(64))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Dropout(rate=0.2))\n",
    "model2.add(Dense(nb_classes))\n",
    "model2.add(Activation('softmax'))\n",
    "model2.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We train the new model with a batch size of 256 for 30 epochs**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:53:13.191472Z",
     "start_time": "2019-10-29T11:45:55.322554Z"
    },
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15290 samples, validate on 3823 samples\n",
      "Epoch 1/30\n",
      "15290/15290 [==============================] - 19s 1ms/step - loss: 761454.4000 - acc: 0.4653 - val_loss: 28340.7632 - val_acc: 0.4311\n",
      "Epoch 2/30\n",
      "15290/15290 [==============================] - 14s 921us/step - loss: 20517.6986 - acc: 0.4780 - val_loss: 15308.5920 - val_acc: 0.4988\n",
      "Epoch 3/30\n",
      "15290/15290 [==============================] - 13s 881us/step - loss: 12844.3107 - acc: 0.4868 - val_loss: 9921.8374 - val_acc: 0.4959\n",
      "Epoch 4/30\n",
      "15290/15290 [==============================] - 15s 1ms/step - loss: 8283.3197 - acc: 0.4942 - val_loss: 6419.6704 - val_acc: 0.4962\n",
      "Epoch 5/30\n",
      "15290/15290 [==============================] - 15s 998us/step - loss: 5339.0282 - acc: 0.4963 - val_loss: 4171.8381 - val_acc: 0.4949\n",
      "Epoch 6/30\n",
      "15290/15290 [==============================] - 14s 944us/step - loss: 3475.9095 - acc: 0.4980 - val_loss: 2737.0189 - val_acc: 0.4949\n",
      "Epoch 7/30\n",
      "15290/15290 [==============================] - 13s 857us/step - loss: 2285.2822 - acc: 0.5184 - val_loss: 1811.1994 - val_acc: 0.5388\n",
      "Epoch 8/30\n",
      "15290/15290 [==============================] - 13s 853us/step - loss: 1514.8119 - acc: 0.5578 - val_loss: 1211.4257 - val_acc: 0.6291\n",
      "Epoch 9/30\n",
      "15290/15290 [==============================] - 14s 912us/step - loss: 1013.3612 - acc: 0.6150 - val_loss: 816.7990 - val_acc: 0.6490\n",
      "Epoch 10/30\n",
      "15290/15290 [==============================] - 13s 862us/step - loss: 682.7646 - acc: 0.6385 - val_loss: 554.2901 - val_acc: 0.6665\n",
      "Epoch 11/30\n",
      "15290/15290 [==============================] - 15s 981us/step - loss: 462.9577 - acc: 0.6523 - val_loss: 378.4063 - val_acc: 0.6746\n",
      "Epoch 12/30\n",
      "15290/15290 [==============================] - 15s 989us/step - loss: 315.8344 - acc: 0.6634 - val_loss: 261.0208 - val_acc: 0.6811\n",
      "Epoch 13/30\n",
      "15290/15290 [==============================] - 15s 975us/step - loss: 217.1697 - acc: 0.6693 - val_loss: 181.1434 - val_acc: 0.6858\n",
      "Epoch 14/30\n",
      "15290/15290 [==============================] - 14s 943us/step - loss: 150.1115 - acc: 0.6730 - val_loss: 128.1626 - val_acc: 0.6898\n",
      "Epoch 15/30\n",
      "15290/15290 [==============================] - 15s 953us/step - loss: 105.2546 - acc: 0.6798 - val_loss: 90.8393 - val_acc: 0.6908\n",
      "Epoch 16/30\n",
      "15290/15290 [==============================] - 15s 958us/step - loss: 74.6190 - acc: 0.6823 - val_loss: 65.4436 - val_acc: 0.6900\n",
      "Epoch 17/30\n",
      "15290/15290 [==============================] - 16s 1ms/step - loss: 53.7518 - acc: 0.6860 - val_loss: 48.2901 - val_acc: 0.6921\n",
      "Epoch 18/30\n",
      "15290/15290 [==============================] - 15s 970us/step - loss: 39.5685 - acc: 0.6948 - val_loss: 36.4650 - val_acc: 0.7015\n",
      "Epoch 19/30\n",
      "15290/15290 [==============================] - 13s 873us/step - loss: 29.8390 - acc: 0.7026 - val_loss: 28.1016 - val_acc: 0.7000\n",
      "Epoch 20/30\n",
      "15290/15290 [==============================] - 13s 882us/step - loss: 23.1054 - acc: 0.7101 - val_loss: 22.2994 - val_acc: 0.7089\n",
      "Epoch 21/30\n",
      "15290/15290 [==============================] - 14s 902us/step - loss: 18.4272 - acc: 0.7302 - val_loss: 18.2831 - val_acc: 0.7162\n",
      "Epoch 22/30\n",
      "15290/15290 [==============================] - 13s 879us/step - loss: 15.1572 - acc: 0.7498 - val_loss: 15.4264 - val_acc: 0.7209\n",
      "Epoch 23/30\n",
      "15290/15290 [==============================] - 13s 877us/step - loss: 12.9085 - acc: 0.7621 - val_loss: 13.6495 - val_acc: 0.6785\n",
      "Epoch 24/30\n",
      "15290/15290 [==============================] - 15s 985us/step - loss: 11.4718 - acc: 0.7715 - val_loss: 12.0752 - val_acc: 0.7298\n",
      "Epoch 25/30\n",
      "15290/15290 [==============================] - 15s 979us/step - loss: 10.5926 - acc: 0.7734 - val_loss: 11.1049 - val_acc: 0.6921\n",
      "Epoch 26/30\n",
      "15290/15290 [==============================] - 16s 1ms/step - loss: 13.4426 - acc: 0.7673 - val_loss: 19.8313 - val_acc: 0.6809\n",
      "Epoch 27/30\n",
      "15290/15290 [==============================] - 14s 907us/step - loss: 11.8941 - acc: 0.7609 - val_loss: 12.8536 - val_acc: 0.7175\n",
      "Epoch 28/30\n",
      "15290/15290 [==============================] - 14s 936us/step - loss: 10.6810 - acc: 0.7821 - val_loss: 9.7520 - val_acc: 0.6974\n",
      "Epoch 29/30\n",
      "15290/15290 [==============================] - 14s 932us/step - loss: 10.3105 - acc: 0.7691 - val_loss: 24.4877 - val_acc: 0.7170\n",
      "Epoch 30/30\n",
      "15290/15290 [==============================] - 14s 933us/step - loss: 18.8480 - acc: 0.7523 - val_loss: 8.6864 - val_acc: 0.6453\n"
     ]
    }
   ],
   "source": [
    "history4 = model2.fit(X_train, y_train, epochs=30, batch_size=256, validation_data=(X_test,y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:58:10.890961Z",
     "start_time": "2019-10-29T11:58:10.372664Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training validation accuracy values\n",
    "plt.plot(history4.history['acc'])\n",
    "plt.plot(history4.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "hidden": true
   },
   "source": [
    "**We reach a worse accuracy, 64.5%, so the higher complexity did NOT provide better results**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-10-29T11:58:48.398275Z",
     "start_time": "2019-10-29T11:58:46.894134Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3823/3823 [==============================] - 1s 369us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.838927321648554, 0.6453047345328824]"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model2.evaluate(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conclusions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**The best result obtained is an accuracy of 74% on the test set which was achieved using a Logistic Regression Model**\n",
    "\n",
    "The results achieved allows us to implement a possible real-time process which, during a game, based on the moves, on the number of turns, on the player ratings etc keeps track of the real-time match outcome probability; when the probability of an outcome goes beyond a certain threshold the platform sends a message to the players suggesting that they should stop the game, thus guaranteeing a high level of entertainment and limiting the boring phases of the game.\n",
    "\n",
    "Moreover, another application would be to implement a game mode where players start to play from a certain pawn configuration, in order to skip the initial boring phase and to start from a point where the outcome of the game is still uncertain. \n",
    "\n",
    "Another game mode could consist of starting from a point where one player, usually the more skilled one, starts from a disadvantaged position, to make the match more challenging. The difficulty level can be determined from the outcome probabilities predicted by the model.\n",
    "\n",
    "\n",
    "\n"
   ]
  }
 ],
 "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": "3.6.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "195px"
   },
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}