Sparkify_Churn_Prediction_Deploy.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sparkify Project Deplpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c08635d3071d445295f2fa843a5b7a37",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting pandas\n",
      "  Using cached https://files.pythonhosted.org/packages/25/47/22fc373440e144e2111363adaa07abb09ec1f03fbc071b6d9fc0bbf65f68/pandas-1.1.3-cp37-cp37m-manylinux1_x86_64.whl\n",
      "Requirement already satisfied: numpy>=1.15.4 in /usr/local/lib64/python3.7/site-packages (from pandas)\n",
      "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/site-packages (from pandas)\n",
      "Collecting python-dateutil>=2.7.3 (from pandas)\n",
      "  Using cached https://files.pythonhosted.org/packages/d4/70/d60450c3dd48ef87586924207ae8907090de0b306af2bce5d134d78615cb/python_dateutil-2.8.1-py2.py3-none-any.whl\n",
      "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas)\n",
      "Installing collected packages: python-dateutil, pandas\n",
      "Successfully installed pandas-1.1.3 python-dateutil-2.8.1\n",
      "\n",
      "Collecting matplotlib\n",
      "  Using cached https://files.pythonhosted.org/packages/87/a6/8d7d06f6b69236a3c1818157875ceb1259ba0d9df4194f4fe138ffdc0f41/matplotlib-3.3.2-cp37-cp37m-manylinux1_x86_64.whl\n",
      "Collecting certifi>=2020.06.20 (from matplotlib)\n",
      "  Using cached https://files.pythonhosted.org/packages/5e/c4/6c4fe722df5343c33226f0b4e0bb042e4dc13483228b4718baf286f86d87/certifi-2020.6.20-py2.py3-none-any.whl\n",
      "Requirement already satisfied: numpy>=1.15 in /usr/local/lib64/python3.7/site-packages (from matplotlib)\n",
      "Requirement already satisfied: python-dateutil>=2.1 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib)\n",
      "Collecting pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.3 (from matplotlib)\n",
      "  Using cached https://files.pythonhosted.org/packages/8a/bb/488841f56197b13700afd5658fc279a2025a39e22449b7cf29864669b15d/pyparsing-2.4.7-py2.py3-none-any.whl\n",
      "Collecting pillow>=6.2.0 (from matplotlib)\n",
      "  Using cached https://files.pythonhosted.org/packages/9f/b7/4b3304e5fd986e2ad8a0157adba88c01dcefc111deaf84a37175af5f5e43/Pillow-8.0.0-cp37-cp37m-manylinux1_x86_64.whl\n",
      "Collecting cycler>=0.10 (from matplotlib)\n",
      "  Using cached https://files.pythonhosted.org/packages/f7/d2/e07d3ebb2bd7af696440ce7e754c59dd546ffe1bbe732c8ab68b9c834e61/cycler-0.10.0-py2.py3-none-any.whl\n",
      "Collecting kiwisolver>=1.0.1 (from matplotlib)\n",
      "  Using cached https://files.pythonhosted.org/packages/31/b9/6202dcae729998a0ade30e80ac00f616542ef445b088ec970d407dfd41c0/kiwisolver-1.2.0-cp37-cp37m-manylinux1_x86_64.whl\n",
      "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/site-packages (from python-dateutil>=2.1->matplotlib)\n",
      "Installing collected packages: certifi, pyparsing, pillow, cycler, kiwisolver, matplotlib\n",
      "Successfully installed certifi-2020.6.20 cycler-0.10.0 kiwisolver-1.2.0 matplotlib-3.3.2 pillow-8.0.0 pyparsing-2.4.7\n",
      "\n",
      "Collecting seaborn\n",
      "  Using cached https://files.pythonhosted.org/packages/bc/45/5118a05b0d61173e6eb12bc5804f0fbb6f196adb0a20e0b16efc2b8e98be/seaborn-0.11.0-py3-none-any.whl\n",
      "Requirement already satisfied: numpy>=1.15 in /usr/local/lib64/python3.7/site-packages (from seaborn)\n",
      "Collecting scipy>=1.0 (from seaborn)\n",
      "  Using cached https://files.pythonhosted.org/packages/65/f9/f7a7e5009711579c72da2725174825e5056741bf4001815d097eef1b2e17/scipy-1.5.2-cp37-cp37m-manylinux1_x86_64.whl\n",
      "Requirement already satisfied: matplotlib>=2.2 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from seaborn)\n",
      "Requirement already satisfied: pandas>=0.23 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from seaborn)\n",
      "Requirement already satisfied: certifi>=2020.06.20 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: python-dateutil>=2.1 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.3 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: pillow>=6.2.0 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: cycler>=0.10 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /mnt/tmp/1602891629876-0/lib/python3.7/site-packages (from matplotlib>=2.2->seaborn)\n",
      "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/site-packages (from pandas>=0.23->seaborn)\n",
      "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/site-packages (from python-dateutil>=2.1->matplotlib>=2.2->seaborn)\n",
      "Installing collected packages: scipy, seaborn\n",
      "Successfully installed scipy-1.5.2 seaborn-0.11.0"
     ]
    }
   ],
   "source": [
    "sc.install_pypi_package(\"pandas\")\n",
    "sc.install_pypi_package(\"matplotlib\")\n",
    "sc.install_pypi_package(\"seaborn\")\n",
    "sc.install_pypi_package(\"sklearn\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fd388dd4aff44c339e97015b0df3a513",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# import libraries\n",
    "from pyspark.sql import SparkSession\n",
    "from pyspark.sql.functions import avg, col, concat, desc, asc, sum, min, max, split, udf,to_date \\\n",
    "                                , concat_ws,first,last,countDistinct,when,count,datediff,trim,split\\\n",
    "                                , element_at ,row_number,explode, array, lit, create_map\n",
    "from pyspark.sql.types import StringType,IntegerType,DoubleType\n",
    "from pyspark.sql import Window\n",
    "\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.classification import LogisticRegression, RandomForestClassifier, GBTClassifier,LinearSVC\n",
    "from pyspark.ml.evaluation import MulticlassClassificationEvaluator,BinaryClassificationEvaluator\n",
    "from pyspark.ml.feature import StandardScaler, StringIndexer,VectorAssembler\n",
    "from pyspark.ml.tuning import CrossValidator, ParamGridBuilder\n",
    "from pyspark.mllib.evaluation import MulticlassMetrics\n",
    "from pyspark.ml.stat import Correlation\n",
    "from itertools import chain\n",
    "import itertools\n",
    "\n",
    "import re\n",
    "\n",
    "import datetime\n",
    "import time\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import warnings; \n",
    "warnings.simplefilter('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Current session configs: <tt>{'conf': {'spark.pyspark.python': 'python3', 'spark.pyspark.virtualenv.enabled': 'true', 'spark.pyspark.virtualenv.type': 'native', 'spark.pyspark.virtualenv.bin.path': '/usr/bin/virtualenv'}, 'kind': 'pyspark'}</tt><br>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table>\n",
       "<tr><th>ID</th><th>YARN Application ID</th><th>Kind</th><th>State</th><th>Spark UI</th><th>Driver log</th><th>Current session?</th></tr><tr><td>1</td><td>application_1602882177189_0002</td><td>pyspark</td><td>idle</td><td><a target=\"_blank\" href=\"http://ip-172-31-45-24.us-east-2.compute.internal:20888/proxy/application_1602882177189_0002/\">Link</a></td><td><a target=\"_blank\" href=\"http://ip-172-31-33-246.us-east-2.compute.internal:8042/node/containerlogs/container_1602882177189_0002_01_000001/livy\">Link</a></td><td>✔</td></tr></table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%info"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "92865577a7b243d8b3322ca8873cd3cb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Package                    Version  \n",
      "-------------------------- ---------\n",
      "beautifulsoup4             4.9.1    \n",
      "boto                       2.49.0   \n",
      "certifi                    2020.6.20\n",
      "click                      7.1.2    \n",
      "cycler                     0.10.0   \n",
      "jmespath                   0.10.0   \n",
      "joblib                     0.16.0   \n",
      "kiwisolver                 1.2.0    \n",
      "lxml                       4.5.2    \n",
      "matplotlib                 3.3.2    \n",
      "mysqlclient                1.4.2    \n",
      "nltk                       3.5      \n",
      "nose                       1.3.4    \n",
      "numpy                      1.16.5   \n",
      "pandas                     1.1.3    \n",
      "Pillow                     8.0.0    \n",
      "pip                        9.0.1    \n",
      "py-dateutil                2.2      \n",
      "pyparsing                  2.4.7    \n",
      "python-dateutil            2.8.1    \n",
      "python37-sagemaker-pyspark 1.4.0    \n",
      "pytz                       2020.1   \n",
      "PyYAML                     5.3.1    \n",
      "regex                      2020.7.14\n",
      "scipy                      1.5.2    \n",
      "seaborn                    0.11.0   \n",
      "setuptools                 28.8.0   \n",
      "six                        1.13.0   \n",
      "soupsieve                  1.9.5    \n",
      "tqdm                       4.48.2   \n",
      "wheel                      0.29.0   \n",
      "windmill                   1.6"
     ]
    }
   ],
   "source": [
    "sc.list_packages()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ee91871600b849588e8d7d42e81fa280",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create spark session\n",
    "spark = SparkSession \\\n",
    "    .builder \\\n",
    "    .appName(\"Sparkify\") \\\n",
    "    .getOrCreate()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load and Clean Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f4b535c28bae496fb512f697a74d6338",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Read in full sparkify dataset\n",
    "event_data = \"s3n://udacity-dsnd/sparkify/sparkify_event_data.json\"\n",
    "df = spark.read.json(event_data)\n",
    "#df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9d844767301943499a1e068595fd4f63",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- artist: string (nullable = true)\n",
      " |-- auth: string (nullable = true)\n",
      " |-- firstName: string (nullable = true)\n",
      " |-- gender: string (nullable = true)\n",
      " |-- itemInSession: long (nullable = true)\n",
      " |-- lastName: string (nullable = true)\n",
      " |-- length: double (nullable = true)\n",
      " |-- level: string (nullable = true)\n",
      " |-- location: string (nullable = true)\n",
      " |-- method: string (nullable = true)\n",
      " |-- page: string (nullable = true)\n",
      " |-- registration: long (nullable = true)\n",
      " |-- sessionId: long (nullable = true)\n",
      " |-- song: string (nullable = true)\n",
      " |-- status: long (nullable = true)\n",
      " |-- ts: long (nullable = true)\n",
      " |-- userAgent: string (nullable = true)\n",
      " |-- userId: string (nullable = true)"
     ]
    }
   ],
   "source": [
    "#print schema\n",
    "df.printSchema()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c2b313290a234f769680930f39c6d3cc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26259199"
     ]
    }
   ],
   "source": [
    "#data size\n",
    "df.count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a9412eb5d8e0429ebc3a12ff708f8be2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------------------+\n",
      "|count(DISTINCT userId)|\n",
      "+----------------------+\n",
      "|                 22278|\n",
      "+----------------------+"
     ]
    }
   ],
   "source": [
    "#Count distinct users\n",
    "df.agg(countDistinct(\"userId\")).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1f3702355f64428daf6bb8ddd5a7c554",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Row(artist=5408927, auth=0, firstName=778479, gender=778479, itemInSession=0, lastName=778479, length=5408927, level=0, location=778479, method=0, page=0, registration=778479, sessionId=0, song=5408927, status=0, ts=0, userAgent=778479, userId=0)]"
     ]
    }
   ],
   "source": [
    "#check missing values for each column\n",
    "df.select(*(sum(col(c).isNull().cast(\"int\")).alias(c) for c in df.columns)).collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f6dada2adb4242948aa647eeb0c1c31d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0"
     ]
    }
   ],
   "source": [
    "#explore users with missing user id\n",
    "df.filter(df.userId == \"\").count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f5200cd7dfc54085966ca8bc80775c77",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "778479"
     ]
    }
   ],
   "source": [
    "#explore users who are either logged out or guests (778479/26259199 = 3% records)\n",
    "df.filter((df.auth == \"Logged Out\")|(df.auth == \"Guest\")).count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bdd1217f894b485a854f78d0e39f7a7f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------------------+\n",
      "|count(DISTINCT userId)|\n",
      "+----------------------+\n",
      "|                     1|\n",
      "+----------------------+"
     ]
    }
   ],
   "source": [
    "#explore users who are either logged out or guests (they share the same userID)\n",
    "df.filter((df.auth == \"Logged Out\")|(df.auth == \"Guest\")).agg(countDistinct(\"userId\")).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8f8c555788c447deaae57d1f10cbd995",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0"
     ]
    }
   ],
   "source": [
    "#explore users with missing sessionId\n",
    "df.filter(df.sessionId == \"\").count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a62ce303afab4a788e335e1bbe19fb4c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#exclude logged out users or Guests\n",
    "df_valid = df.filter((df.auth != \"Logged Out\")&(df.auth != \"Guest\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Engineering"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define Churn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f7bcd1b271f249a7a0248e84521bfae3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#flag Cancellation Confirmation event\n",
    "flag_cancel_event = udf(lambda x: 1 if x == \"Cancellation Confirmation\" else 0, IntegerType())\n",
    "df_valid = df_valid.withColumn(\"Cancellation\", flag_cancel_event(\"page\"))\n",
    "df_valid.filter(df_valid.page =='Cancellation Confirmation').head()\n",
    "#define Churn users\n",
    "windowval = Window.partitionBy(\"userId\")\n",
    "df_valid = df_valid.withColumn(\"Churn\", max(\"Cancellation\").over(windowval))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0a25ceaf40a345d7a8652465a9508064",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#add date time\n",
    "get_year = udf(lambda x: datetime.datetime.fromtimestamp(x / 1000.0).year)\n",
    "get_month = udf(lambda x: datetime.datetime.fromtimestamp(x / 1000.0).month)\n",
    "get_day = udf(lambda x: datetime.datetime.fromtimestamp(x / 1000.0).day)\n",
    "get_hour = udf(lambda x: datetime.datetime.fromtimestamp(x / 1000.0).hour)\n",
    "df_valid = df_valid.withColumn(\"year\",get_year(\"ts\"))\n",
    "df_valid = df_valid.withColumn(\"month\",get_month(\"ts\"))\n",
    "df_valid = df_valid.withColumn(\"day\",get_day(\"ts\"))\n",
    "df_valid = df_valid.withColumn(\"hour\",get_hour(\"ts\"))\n",
    "df_valid = df_valid.withColumn(\"date\", to_date(concat_ws(\"-\", \"year\", \"month\", \"day\")))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "656c5e922e41448996cbae7fdb0b9e01",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+----------+----------+\n",
      "| max(date)| min(date)|\n",
      "+----------+----------+\n",
      "|2018-12-01|2018-10-01|\n",
      "+----------+----------+"
     ]
    }
   ],
   "source": [
    "#check last data and first date \n",
    "df_valid.agg(max(col('date')),min(col('date'))).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ef2841bdb37d47659ba98dc901cef58d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# main state (highest frequency)\n",
    "df_valid = df_valid.withColumn(\"state\", trim(element_at(split(col(\"location\"), \",\"),-1)))\n",
    "state_cnt = df_valid.groupBy(\"userId\", \"state\").count()\n",
    "\n",
    "w = Window().partitionBy(\"userId\").orderBy(col(\"count\").desc())\n",
    "\n",
    "main_state = (state_cnt\n",
    "  .withColumn(\"rn\", row_number().over(w))\n",
    "  .where(col(\"rn\") == 1)\n",
    "  .select(\"userId\", col(\"state\").alias('main_state')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "09c1f043875241068caa47832866947f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Adding main state feature to df_valid\n",
    "df_valid = df_valid.join(main_state, on = 'userId')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a1803911922c4e75b5424e525965ce01",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Extract potential features \n",
    "df_user = df_valid.groupby('userId').agg(\n",
    "     # target\n",
    "     first(col('Churn')).alias('label'),\n",
    "     # gender\n",
    "     first(when(col('gender') == \"F\", 1).otherwise(0)).alias('female'),\n",
    "     # #main state ratio\n",
    "     #((sum(when(col('state') == col('main_state'), 1).otherwise(0)))/count(col('state'))).alias(\"mstate_ratio\"),\n",
    "     # distinct locations\n",
    "     #countDistinct(col('location')).alias('ulocation_ct'),\n",
    "     # distinct page actions\n",
    "     countDistinct(col('page')).alias('upage_ct'),\n",
    "     # page action ratio\n",
    "    ((sum(when(col('page') == \"Thumbs Up\", 1).otherwise(0)))/count(col('page'))).alias(\"tbup_ratio\"),\n",
    "    ((sum(when(col('page') == \"Thumbs Down\", 1).otherwise(0)))/count(col('page'))).alias(\"tbdw_ratio\"),\n",
    "    ((sum(when(col('page') == \"Add Friend\", 1).otherwise(0)))/count(col('page'))).alias(\"addfriend_ratio\"),   \n",
    "    ((sum(when(col('page') == \"Roll Advert\", 1).otherwise(0)))/count(col('page'))).alias(\"ad_ratio\"),\n",
    "    ((sum(when(col('page') == \"Upgrade\", 1).otherwise(0)))/count(col('page'))).alias(\"up_ratio\"),\n",
    "    ((sum(when(col('page') == \"Help\", 1).otherwise(0)))/count(col('page'))).alias(\"help_ratio\"),\n",
    "    ((sum(when(col('page') == \"Error\", 1).otherwise(0)))/count(col('page'))).alias(\"error_ratio\"),\n",
    "     # enroll days\n",
    "     (datediff(max(col('date')),min(col('date'))) +1).alias(\"enroll_days\"),\n",
    "     # active days\n",
    "     countDistinct(col('date')).alias('active_days'),\n",
    "     # active ratio\n",
    "     (countDistinct(col('date'))/(datediff(max(col('date')),min(col('date'))) +1)).alias('active_ratio'),\n",
    "     # distinct artists per active day \n",
    "     (countDistinct(col('artist'))/countDistinct(col('date'))).alias('uartist_pact'),\n",
    "     # distinct songs per active day\n",
    "     (countDistinct(col('song'))/countDistinct(col('date'))).alias('usong_pact'),\n",
    "     # avg songs per active day \n",
    "     (sum(when(col('page') == \"NextSong\",1).otherwise(0))/countDistinct(col('date'))).alias(\"tsong_pact\"),\n",
    "     # avg items per session\n",
    "     (count(col('itemInSession'))/countDistinct(col('sessionId'))).alias(\"item_pses\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bb9ebe41d0dc48e1a7ffd3de42280435",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Row(userId='1076191', label=1, female=0, mstate_ratio=1.0, ulocation_ct=1, upage_ct=9, tbup_ratio=0.0625, tbdw_ratio=0.015625, addfriend_ratio=0.0, ad_ratio=0.078125, up_ratio=0.0, help_ratio=0.0, error_ratio=0.0, enroll_days=9, active_days=3, active_ratio=0.3333333333333333, uartist_pact=15.666666666666666, usong_pact=15.666666666666666, tsong_pact=15.666666666666666, item_pses=21.333333333333332)"
     ]
    }
   ],
   "source": [
    "#check user example\n",
    "df_user.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Explore Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "78d3cb5bca5a47ffac39d1637f8c47e8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- userId: string (nullable = true)\n",
      " |-- label: integer (nullable = true)\n",
      " |-- female: integer (nullable = true)\n",
      " |-- mstate_ratio: double (nullable = true)\n",
      " |-- ulocation_ct: long (nullable = false)\n",
      " |-- upage_ct: long (nullable = false)\n",
      " |-- tbup_ratio: double (nullable = true)\n",
      " |-- tbdw_ratio: double (nullable = true)\n",
      " |-- addfriend_ratio: double (nullable = true)\n",
      " |-- ad_ratio: double (nullable = true)\n",
      " |-- up_ratio: double (nullable = true)\n",
      " |-- help_ratio: double (nullable = true)\n",
      " |-- error_ratio: double (nullable = true)\n",
      " |-- enroll_days: integer (nullable = true)\n",
      " |-- active_days: long (nullable = false)\n",
      " |-- active_ratio: double (nullable = true)\n",
      " |-- uartist_pact: double (nullable = true)\n",
      " |-- usong_pact: double (nullable = true)\n",
      " |-- tsong_pact: double (nullable = true)\n",
      " |-- item_pses: double (nullable = true)"
     ]
    }
   ],
   "source": [
    "#print schema\n",
    "df_user.printSchema()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### #target measure plots "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fa38e596864d41c8998056577f23a1ce",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  label  count   Percent\n",
      "0     1   5003  0.224581\n",
      "1     0  17274  0.775419"
     ]
    }
   ],
   "source": [
    "#prepare target data\n",
    "target_df = df_user.groupby('label').count()\n",
    "target_pdf = target_df.toPandas()\n",
    "target_pdf['label'] = target_pdf['label'].astype('category')\n",
    "target_pdf['Percent'] = target_pdf['count']/target_pdf['count'].sum() \n",
    "target_pdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0b0ebab4650a4376874eef7d0383dc32",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Text(-0.05769230769230771, 0.23458140683215875, '22.0%')\n",
      "Text(0.9423076923076923, 0.7854185931678412, '78.0%')"
     ]
    }
   ],
   "source": [
    "#target plot\n",
    "churn_pt = target_pdf.plot(kind ='bar',x ='label',y='Percent',legend = None,color ='navy', \\\n",
    "            title = 'Churn Distribution (Full 12GB Dataset)',xlabel = '', ylabel ='Percent', figsize = (8,5));\n",
    "for p in churn_pt.patches:\n",
    "    txt = str(p.get_height().round(2)* 100) + '%'\n",
    "    txt_x = p.get_x() + p.get_width() / 2.6\n",
    "    txt_y = p.get_height()+0.01\n",
    "    churn_pt.text(txt_x,txt_y,txt)\n",
    "#plt.savefig('s3://aws-emr-resources-631922983932-us-east-2/e-C01H7UNHA5J04BSQL7GOJGG6X/churn.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#img_data = io.BytesIO()\n",
    "#plt.savefig(img_data, format='png')\n",
    "#img_data.seek(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#####  fields with only one distinct value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "377250977e0641eb87f4f2ddd1794226",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+------+-----+------+------------+------------+--------+----------+----------+---------------+--------+--------+----------+-----------+-----------+-----------+------------+------------+----------+----------+---------+\n",
      "|userId|label|female|mstate_ratio|ulocation_ct|upage_ct|tbup_ratio|tbdw_ratio|addfriend_ratio|ad_ratio|up_ratio|help_ratio|error_ratio|enroll_days|active_days|active_ratio|uartist_pact|usong_pact|tsong_pact|item_pses|\n",
      "+------+-----+------+------------+------------+--------+----------+----------+---------------+--------+--------+----------+-----------+-----------+-----------+------------+------------+----------+----------+---------+\n",
      "| 22277|    2|     2|           1|           1|      19|     11880|      9465|          10383|   12897|    7078|      7913|       5304|         61|         61|         958|        8394|      9220|      9563|    10456|\n",
      "+------+-----+------+------------+------------+--------+----------+----------+---------------+--------+--------+----------+-----------+-----------+-----------+------------+------------+----------+----------+---------+"
     ]
    }
   ],
   "source": [
    "# countDistinct value for each column\n",
    "expr = [countDistinct(c).alias(c) for c in df_user.columns]\n",
    "\n",
    "# apply those functions\n",
    "countdf =  df_user.select(*expr)\n",
    "countdf.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3c0eb09bb3ef49beb775c60d157d42bc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['mstate_ratio', 'ulocation_ct']"
     ]
    }
   ],
   "source": [
    "# extract the columns which have just one value\n",
    "one_value_to_drop = [k for k,v in countdf.collect()[0].asDict().items() if v == 1]\n",
    "one_value_to_drop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### fields with high correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6defa67b141d4270a5465370e8da49d7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['error_ratio', 'usong_pact', 'upage_ct', 'addfriend_ratio', 'active_days', 'enroll_days', 'item_pses', 'tbdw_ratio', 'up_ratio', 'ad_ratio', 'active_ratio', 'uartist_pact', 'tsong_pact', 'label', 'help_ratio', 'tbup_ratio']"
     ]
    }
   ],
   "source": [
    "#numeric variables\n",
    "num_vector = list(set(df_user.columns)-set(['userId','female'])-set(one_value_to_drop))\n",
    "num_vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7bb9c55962784d5c8afe7be188d0249f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# convert to vector column first\n",
    "vector_col = \"corr_features\"\n",
    "assembler = VectorAssembler(inputCols=df_user.select(num_vector).columns, outputCol=vector_col)\n",
    "df_vector = assembler.transform(df_user.select(num_vector)).select(vector_col)\n",
    "\n",
    "# get correlation matrix\n",
    "matrix = Correlation.corr(df_vector, vector_col)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "22caeff1d4544080be18646207398c23",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#matrix head name\n",
    "head = df_user.select(num_vector).columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ed99795b65f34c3b9951c2619ca2c0d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                 error_ratio  usong_pact  ...  help_ratio  tbup_ratio\n",
      "error_ratio         1.000000   -0.017840  ...   -0.014701   -0.040267\n",
      "usong_pact         -0.017840    1.000000  ...   -0.010930    0.101100\n",
      "upage_ct            0.040470    0.534562  ...    0.066610    0.075630\n",
      "addfriend_ratio    -0.013183   -0.014889  ...    0.028843   -0.069015\n",
      "active_days        -0.013901    0.304876  ...    0.007492    0.118906\n",
      "enroll_days        -0.017652    0.080075  ...    0.030309    0.130688\n",
      "item_pses          -0.014321    0.904392  ...   -0.007082    0.089550\n",
      "tbdw_ratio         -0.006346   -0.082510  ...   -0.004458   -0.100777\n",
      "up_ratio           -0.009756   -0.302070  ...   -0.003553   -0.072787\n",
      "ad_ratio            0.006918   -0.526481  ...   -0.010654   -0.280387\n",
      "active_ratio        0.009217    0.223136  ...   -0.039895   -0.039150\n",
      "uartist_pact       -0.016764    0.972071  ...   -0.012758    0.088603\n",
      "tsong_pact         -0.018012    0.992287  ...   -0.010030    0.104174\n",
      "label              -0.001909    0.124695  ...   -0.015775   -0.131076\n",
      "help_ratio         -0.014701   -0.010930  ...    1.000000   -0.027032\n",
      "tbup_ratio         -0.040267    0.101100  ...   -0.027032    1.000000\n",
      "\n",
      "[16 rows x 16 columns]"
     ]
    }
   ],
   "source": [
    "#correlation matrix\n",
    "corr = pd.DataFrame(matrix.collect()[0][\"pearson({})\".format(vector_col)].values.reshape(16,16))\n",
    "corr.columns = head\n",
    "corr.index = head\n",
    "corr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "296b592d61984b2aa38e17830a6585fc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['uartist_pact', 'tsong_pact']"
     ]
    }
   ],
   "source": [
    "# Creating ABS correlation matrix\n",
    "corr_abs = corr.abs()\n",
    "# Selecting upper triangle of correlation matrix\n",
    "upper_tri = corr_abs.where(np.triu(np.ones(corr_abs.shape),k=1).astype(np.bool))\n",
    "# Finding feature columns with correlation greater than 0.95\n",
    "high_cor_to_drop = [column for column in upper_tri.columns if any(upper_tri[column] > 0.95)]\n",
    "high_cor_to_drop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### drop fields from two lists above"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6f49d95650334f62aa8299174bf95ee2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['mstate_ratio', 'ulocation_ct', 'uartist_pact', 'tsong_pact']"
     ]
    }
   ],
   "source": [
    "# transfer the drop list to spark data frame\n",
    "drop_list = one_value_to_drop + high_cor_to_drop\n",
    "drop_list "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d6a674e9638044c7ba1809bcf9c7c516",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#drop_list = ['uartist_pact', 'tsong_pact']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "430e6741ad3344318ca7dff7e4db51fb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#drop fields from drop list\n",
    "df_user_clean = df_user.drop(*drop_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f41195706546486898109053bf44d872",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['userId', 'label', 'female', 'upage_ct', 'tbup_ratio', 'tbdw_ratio', 'addfriend_ratio', 'ad_ratio', 'up_ratio', 'help_ratio', 'error_ratio', 'enroll_days', 'active_days', 'active_ratio', 'usong_pact', 'item_pses']"
     ]
    }
   ],
   "source": [
    "#check final data frame\n",
    "df_user_clean.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Modelling and Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 1. Split the full dataset into train (cross-validation) and test sets. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2bb0a910f653461880cffa3ee5300074",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#split the data\n",
    "train, test = df_user_clean.randomSplit([0.7, 0.3], seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e62670f156844469ad83f2b94cb144fb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+\n",
      "|label|count|\n",
      "+-----+-----+\n",
      "|    1| 3448|\n",
      "|    0|12128|\n",
      "+-----+-----+"
     ]
    }
   ],
   "source": [
    "#check churn distribution\n",
    "train.groupby('label').count().show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 2. Build machine learning Pipeline."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7e151f8c8eff4b4c94d4d78a19a191f6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#create initial machine learning pielines\n",
    "def initial_model(numcols,catcols,classifier = LogisticRegression(maxIter=10),train = train, test = test):\n",
    "    '''\n",
    "    INPUT:\n",
    "    numcols - numerical columns\n",
    "    catcols - categorical columns (binary)\n",
    "    classifier - machine learning classifier model\n",
    "    train - train data\n",
    "    test - test data\n",
    "    \n",
    "    OUTPUT:\n",
    "    pipeline - machine learning pipeline\n",
    "    Model - fitted model based on train data\n",
    "    prediction - predictions on test data\n",
    "    run time in mins\n",
    "    '''\n",
    "    \n",
    "    start = time.time()\n",
    "    # Combining into VectorAssembler\n",
    "    num_assembler = VectorAssembler(inputCols = numcols, outputCol = \"num_vector\")\n",
    "    # Standardizing numerical features\n",
    "    scaler = StandardScaler(inputCol = \"num_vector\", outputCol = \"num_scaled\", withStd = True, withMean = True)\n",
    "\n",
    "    # Add binary feature (do not scale)\n",
    "    total_assembler = VectorAssembler(inputCols = catcols + [\"num_scaled\"], outputCol = \"features\")\n",
    "\n",
    "    # choose classifiers\n",
    "    ml =  classifier\n",
    "    pipeline = Pipeline(stages = [num_assembler, scaler, total_assembler, ml])\n",
    "    Model = pipeline.fit(train)\n",
    "    prediction = Model.transform(test) \n",
    "    \n",
    "    end = time.time()\n",
    "    \n",
    "    return pipeline,Model, prediction,round(round(end-start)/60,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4fcfa962798d417ea2429fda9b27ee09",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#evaluation function\n",
    "def evaluation(prediction):\n",
    "    '''\n",
    "    INPUT:\n",
    "    prediction - predictions on test data\n",
    "    \n",
    "    OUTPUT:\n",
    "    accuracy - metric: (TP + TN)/(TP + TN + FP + FN)\n",
    "    F1 score - metric: 2 * (Recall * Precision) / (Recall + Precision)\n",
    "    '''\n",
    "    \n",
    "    # Input: model prediction results\n",
    "    # Output: accuracy and F1 score\n",
    "       # accuracy: (TP + TN)/(TP + TN + FP + FN)\n",
    "       # precision: TP/(TP + FP)\n",
    "       # recall: TP/(TP + FN)\n",
    "       # f1 score: 2*(Recall * Precision) / (Recall + Precision)\n",
    "  \n",
    "    # Combine prediction and labels\n",
    "    predictionAndLabels = prediction.select(['prediction','label']) \\\n",
    "                      .withColumn(\"label\",col(\"label\").cast(DoubleType())).rdd\n",
    "\n",
    "    # Instantiate metrics object\n",
    "    metrics = MulticlassMetrics(predictionAndLabels)\n",
    "    \n",
    "    # Overall statistics\n",
    "    accuracy = metrics.accuracy\n",
    "    precision = metrics.precision(label =1.0)\n",
    "    recall = metrics.recall(label =1.0)\n",
    "    f1Score = metrics.fMeasure(label =1.0)\n",
    "    print(\"Summary Stats\")\n",
    "    print(\"Accuracy = %s\" % accuracy)\n",
    "    print(\"Precision = %s\" % precision)\n",
    "    print(\"Recall = %s\" % recall)\n",
    "    print(\"F1 Score = %s\" % f1Score)\n",
    "       \n",
    "    return round(accuracy,3),round(f1Score,3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6e0be050a5c54c35ad68e35e8338bb9b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['error_ratio', 'usong_pact', 'upage_ct', 'addfriend_ratio', 'item_pses', 'active_days', 'enroll_days', 'tbdw_ratio', 'up_ratio', 'ad_ratio', 'active_ratio', 'help_ratio', 'tbup_ratio']"
     ]
    }
   ],
   "source": [
    "#final features used in the model\n",
    "numcols = list(set(df_user_clean.columns)-set(['userId','female','label']))\n",
    "catcols = ['female']\n",
    "numcols"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "60be42ae62a74249b1f86ac92abf8d7f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Run time (mins):  5.52\n",
      "Summary Stats\n",
      "Accuracy = 0.9195642441426652\n",
      "Precision = 0.8724340175953079\n",
      "Recall = 0.7652733118971061\n",
      "F1 Score = 0.8153477218225419"
     ]
    }
   ],
   "source": [
    "# Create logistic regression classifier\n",
    "pipeline_lr, model_lr, prediction_lr, time_lr = initial_model(numcols,catcols, \\\n",
    "                                                              LogisticRegression(maxIter=10),train, test)\n",
    "print('Logistic Regression Run time (mins): ',time_lr)\n",
    "acc_lr, f1_lr = evaluation(prediction_lr)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "aeac13d8ada548b19089c0815b22d693",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random Forest Run time (mins):  17.15\n",
      "Summary Stats\n",
      "Accuracy = 0.9068795702134009\n",
      "Precision = 0.9251141552511416\n",
      "Recall = 0.6514469453376206\n",
      "F1 Score = 0.7645283018867925"
     ]
    }
   ],
   "source": [
    "# Create Random Forest classifier \n",
    "pipeline_rf, model_rf, prediction_rf, time_rf = initial_model(numcols,catcols, \\\n",
    "                                                              RandomForestClassifier(),train, test)\n",
    "print('Random Forest Run time (mins): ',time_rf)\n",
    "acc_rf, f1_rf = evaluation(prediction_rf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Gradient-boosted Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f7d0ae3d1360407ab29701630268ee17",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient-boosted Tree Run time (mins):  6.02\n",
      "Summary Stats\n",
      "Accuracy = 0.933293538277869\n",
      "Precision = 0.9314641744548287\n",
      "Recall = 0.7691318327974277\n",
      "F1 Score = 0.8425501937301867"
     ]
    }
   ],
   "source": [
    "# Create Gradient-boosted tree classifier \n",
    "pipeline_gbt, model_gbt, prediction_gbt, time_gbt = initial_model(numcols,catcols, \\\n",
    "                                                                  GBTClassifier(maxIter=10),train, test)\n",
    "print('Gradient-boosted Tree Run time (mins): ',time_gbt)\n",
    "acc_gbt, f1_gbt = evaluation(prediction_gbt)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Linear Support Vector Machine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a78caf6da8944bf39a9aec19125524c5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear Support Vector Machine Run time (mins):  14.88\n",
      "Summary Stats\n",
      "Accuracy = 0.9185196239367258\n",
      "Precision = 0.8859984697781178\n",
      "Recall = 0.7446945337620579\n",
      "F1 Score = 0.8092243186582809"
     ]
    }
   ],
   "source": [
    "# Create Linear Support Vector Machine classifier \n",
    "pipeline_svc, model_svc, prediction_svc, time_svc = initial_model(numcols,catcols, \\\n",
    "                                                                  LinearSVC(maxIter=10),train, test)\n",
    "print('Linear Support Vector Machine Run time (mins): ',time_svc) \n",
    "acc_svc, f1_svc = evaluation(prediction_svc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Initial Models Performance Summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0e769a128d054861a43bd6c4309579ff",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Performance Summary for initial models\n",
      "\n",
      "Logistic Regression model results: F1 Score 0.815, Accuracy is 0.92\n",
      "Random Forest model results: F1 Score 0.765, Accuracy is 0.907\n",
      "Gradient-boosted Tree model results: F1 Score 0.843, Accuracy is 0.933\n",
      "Linear Support Vector Machine results: F1 Score 0.809, Accuracy is 0.919"
     ]
    }
   ],
   "source": [
    "# Model Performance summary for initial models\n",
    "print('Performance Summary for initial models\\n')\n",
    "print(\"Logistic Regression model results: F1 Score {}, Accuracy is {}\".format(f1_lr, acc_lr))\n",
    "print(\"Random Forest model results: F1 Score {}, Accuracy is {}\".format(f1_rf, acc_rf))\n",
    "print(\"Gradient-boosted Tree model results: F1 Score {}, Accuracy is {}\".format(f1_gbt, acc_gbt))\n",
    "print(\"Linear Support Vector Machine results: F1 Score {}, Accuracy is {}\".format(f1_svc, acc_svc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "| Initial Models | F1 Score | Accuracy |\n",
    "| --- | --- | --- |\n",
    "| Logistic Regression | 0.815 | 0.920|\n",
    "| Random Forest | 0.765| 0.907|\n",
    "| Gradient-boosted Tree | 0.843 | 0.933|\n",
    "| Linear Support Vector Machine | 0.809 | 0.919|"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 3. Implement three methods to treat imbalanced class issue."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The data have imblanced classes issue. So I try 3 different methods below to handle this.\n",
    "1. Oversampling\n",
    "2. Undersampling\n",
    "3. Balancing class weights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Stratified split dataframes "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4a01bd7d45e14cb3a068ff2c102063cf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# seperate minority/majority data\n",
    "label1_full = df_user_clean.filter(df_user_clean.label == 1)\n",
    "label0_full = df_user_clean.filter(df_user_clean.label == 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# calculate unbalanced ratio\n",
    "#ratio = label0_full.count()/ label1_full.count()\n",
    "#ratio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9bc9af68375e44fb9c3eb26cee10f134",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# split datasets into training and testing\n",
    "train0, test0 = label0_full.randomSplit([0.7,0.3], seed=42)\n",
    "train1, test1 = label1_full.randomSplit([0.7,0.3], seed=42)\n",
    "# stack datasets back together\n",
    "train_stra = train0.union(train1)\n",
    "test_stra = test0.union(test1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "655d319e7f5543b9ad68cc3524e7a9f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# seperate minority/majority on training data\n",
    "label1_df = train_stra.filter(train_stra.label == 1)\n",
    "label0_df = train_stra.filter(train_stra.label == 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d2a0193ef2a14931839b237129a220eb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Row(label=1, count=3495), Row(label=0, count=12068)]"
     ]
    }
   ],
   "source": [
    "#sample size before oversampling \n",
    "y = train_stra.groupby('label').count().collect()\n",
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7028de8716e34be18aab63cca42bf516",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3.4529327610872675"
     ]
    }
   ],
   "source": [
    "# calculate unbalanced ratio\n",
    "ratio = y[1][1]/ y[0][1]\n",
    "ratio"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1. Oversampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a888087380b14fb795b756ea9ec80927",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# duplicate the minority data (label1)\n",
    "label1_os = label1_df.withColumn(\"dummy\", explode(array([lit(x) for x in range(int(ratio))]))).drop('dummy')\n",
    "# combine both oversampled minority label 1 and previous majority label 0\n",
    "combined_os_df = label0_df.unionAll(label1_os)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0e05286753e9499fb87115e4ce8b9484",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Run time (mins):  5.97\n",
      "Summary Stats\n",
      "Accuracy = 0.9070598748882931\n",
      "Precision = 0.7517084282460137\n",
      "Recall = 0.8753315649867374\n",
      "F1 Score = 0.8088235294117646"
     ]
    }
   ],
   "source": [
    "# Create logistic regression classifier\n",
    "pipeline_lr_os, model_lr_os, prediction_lr_os, time_lr_os = initial_model(numcols,catcols, \\\n",
    "                                                              LogisticRegression(maxIter=10), \\\n",
    "                                                              combined_os_df, test_stra)\n",
    "print('Logistic Regression Run time (mins): ',time_lr_os)\n",
    "acc_lr_os, f1_lr_os = evaluation(prediction_lr_os)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2a1ac7606ec74f759c2b50bbcd8753cd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient-boosted tree Run time (mins):  6.37\n",
      "Summary Stats\n",
      "Accuracy = 0.9139112302651177\n",
      "Precision = 0.7600671140939598\n",
      "Recall = 0.9011936339522546\n",
      "F1 Score = 0.8246359223300971"
     ]
    }
   ],
   "source": [
    "# Create Gradient-boosted tree classifier\n",
    "pipeline_gbt_os, model_gbt_os, prediction_gbt_os, time_gbt_os = initial_model(numcols,catcols, \\\n",
    "                                                              GBTClassifier(maxIter=10), \\\n",
    "                                                              combined_os_df, test_stra)\n",
    "print('Gradient-boosted tree Run time (mins): ',time_gbt_os)\n",
    "acc_gbt_os, f1_gbt_os = evaluation(prediction_gbt_os)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. Undersampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "753b4a84b43b44c5bd96d1adcf2f7994",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#random delete some records from majority label 0\n",
    "label0_us  = label0_df.sample(False, 1/ratio)\n",
    "# combine both undersampled majority label 0 and previous minority label 1\n",
    "combined_us_df = label1_df.unionAll(label0_us)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bfb49c8a11c24696addc85932d6ddaf7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+-----+-----+\n",
      "|label|count|\n",
      "+-----+-----+\n",
      "|    1| 3469|\n",
      "|    0| 3529|\n",
      "+-----+-----+"
     ]
    }
   ],
   "source": [
    "#sample size after undersampling \n",
    "combined_us_df.groupby('label').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "558df4b7bd7b44119d2adff4e777f70f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Run time (mins):  15.77\n",
      "Summary Stats\n",
      "Accuracy = 0.8985701519213584\n",
      "Precision = 0.7250952640174197\n",
      "Recall = 0.883289124668435\n",
      "F1 Score = 0.7964125560538118"
     ]
    }
   ],
   "source": [
    "# Create logistic regression classifier\n",
    "pipeline_lr_us, model_lr_us, prediction_lr_us, time_lr_us = initial_model(numcols,catcols, \\\n",
    "                                                              LogisticRegression(maxIter=10), \\\n",
    "                                                              combined_us_df, test_stra)\n",
    "print('Logistic Regression Run time (mins): ',time_lr_us)\n",
    "acc_lr_us, f1_lr_us = evaluation(prediction_lr_us)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "93e8d88b4e5344018f435fd13a1e872f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient-boosted tree Run time (mins):  7.1\n",
      "Summary Stats\n",
      "Accuracy = 0.9100387250521299\n",
      "Precision = 0.7475355969331873\n",
      "Recall = 0.9051724137931034\n",
      "F1 Score = 0.8188362327534494"
     ]
    }
   ],
   "source": [
    "# Create Gradient-boosted tree classifier\n",
    "pipeline_gbt_us, model_gbt_us, prediction_gbt_us, time_gbt_us = initial_model(numcols,catcols, \\\n",
    "                                                              GBTClassifier(maxIter=10), \\\n",
    "                                                              combined_us_df, test_stra)\n",
    "print('Gradient-boosted tree Run time (mins): ',time_gbt_us)\n",
    "acc_gbt_us, f1_gbt_us = evaluation(prediction_gbt_us)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Balancing class weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#REFER: https://danvatterott.com/blog/2019/11/18/balancing-model-weights-in-pyspark/\n",
    "#calculate class weights in pandas data frame\n",
    "# https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html\n",
    "# class weight: n_samples / (n_classes * np.bincount(y))."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "97218df7558942e8bfe7a5bb25616602",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{1: 2.2264663805436338, 0: 0.644804441498177}"
     ]
    }
   ],
   "source": [
    "#calculate class weights in spark data frame\n",
    "y = train_stra.groupby('label').count().collect()\n",
    "unique_y = [x[\"label\"] for x in y]\n",
    "total_y = train_stra.count()\n",
    "unique_y_count = len(unique_y)\n",
    "bin_count = [x[\"count\"] for x in y]\n",
    "\n",
    "class_weights_spark = {i: ii for i, ii in zip(unique_y, total_y / (unique_y_count * np.array(bin_count)))}\n",
    "print(class_weights_spark) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "30df4034ab1a4e57a64e24c8f59dfe4f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#create a mapping to apply a weight to each training instance.\n",
    "mapping_weight = create_map([lit(x) for x in chain(*class_weights_spark.items())])\n",
    "train_stra_weight = train_stra.withColumn(\"weight\", mapping_weight.getItem(col(\"label\")))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "19e24a99e21d422a82c7b7d77cc702f3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Run time (mins):  9.15\n",
      "Summary Stats\n",
      "Accuracy = 0.8994638069705094\n",
      "Precision = 0.7279693486590039\n",
      "Recall = 0.8819628647214854\n",
      "F1 Score = 0.7976011994002998"
     ]
    }
   ],
   "source": [
    "# Create logistic regression classifier\n",
    "pipeline_lr_wc, model_lr_wc, prediction_lr_wc, time_lr_wc = initial_model(numcols,catcols, \\\n",
    "                                                              LogisticRegression(maxIter=10,weightCol=\"weight\"), \\\n",
    "                                                              train_stra_weight, test_stra)\n",
    "print('Logistic Regression Run time (mins): ',time_lr_wc)\n",
    "acc_lr_wc, f1_lr_wc = evaluation(prediction_lr_wc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Fixing Imbalanced Classes Performance Summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model Performance summary for fixing imbalanced classes\n",
    "print('Performance Summary for Fixing Imbalanced Classes\\n')\n",
    "print('Logistic Regression model')\n",
    "print(\"Initial Logistic Regression model results: F1 Score {}, Accuracy is {}\".format(f1_lr, acc_lr))\n",
    "print(\"1. Oversampling: F1 Score {}, Accuracy is {}\".format(f1_lr_os, acc_lr_os))\n",
    "print(\"2. Undersampling: F1 Score {}, Accuracy is {}\".format(f1_lr_us, acc_lr_us))\n",
    "print(\"3. Balancing class weights: F1 Score {}, Accuracy is {}\".format(f1_lr_wc, acc_lr_wc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "| Category | F1 Score | Accuracy |\n",
    "| --- | --- | --- |\n",
    "| Initial Logistic Regression | 0.815 | 0.920|\n",
    "| Over-sampling | 0.809 | 0.907|\n",
    "| Under-sampling | 0.796 | 0.899|\n",
    "| Balancing class weights | 0.798 | 0.899|"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ed0f200c78db410dba2e4425cef0deae",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient-boosted Tree model\n",
      "Initial Gradient-boosted Tree model results: F1 Score 0.843, Accuracy is 0.933\n",
      "1. Oversampling: F1 Score 0.825, Accuracy is 0.914\n",
      "2. Undersampling: F1 Score 0.819, Accuracy is 0.91"
     ]
    }
   ],
   "source": [
    "# Model Performance summary for fixing imbalanced classes\n",
    "print('Gradient-boosted Tree model')\n",
    "print(\"Initial Gradient-boosted Tree model results: F1 Score {}, Accuracy is {}\".format(f1_gbt, acc_gbt))\n",
    "print(\"1. Oversampling: F1 Score {}, Accuracy is {}\".format(f1_gbt_os, acc_gbt_os))\n",
    "print(\"2. Undersampling: F1 Score {}, Accuracy is {}\".format(f1_gbt_us, acc_gbt_us))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "| Category | F1 Score | Accuracy |\n",
    "| --- | --- | --- |\n",
    "| Initial Gradient-boosted Tree | 0.843 | 0.933|\n",
    "| Over-sampling | 0.825 | 0.914|\n",
    "| Under-sampling | 0.819 | 0.910|"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 4. Conduct 3-fold cross validation and tune parameters. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "79ac24bdc25d42c1b534cf223a047f1a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# cross validation and tune models\n",
    "def model_tuning(pipeline,paramGrid,train = train, test = test):\n",
    "    '''\n",
    "    INPUT:\n",
    "    pipeline - machine learning pipeline\n",
    "    paramGrid - parameter grid for grid search\n",
    "    train - train data\n",
    "    test - test data\n",
    "    \n",
    "    OUTPUT:\n",
    "    Model - fitted model based on train data\n",
    "    prediction - predictions on test data\n",
    "    run time in mins\n",
    "    '''\n",
    "    start = time.time()\n",
    "    # Combining into VectorAssembler\n",
    "    crossval = CrossValidator(estimator=pipeline,\n",
    "                          estimatorParamMaps=paramGrid,\n",
    "                          evaluator=MulticlassClassificationEvaluator(metricName='f1'),\n",
    "                          numFolds=3)\n",
    "    cvmodel = crossval.fit(train)\n",
    "    #predict on test data\n",
    "    cvprediction = cvmodel.transform(test)\n",
    "    end = time.time()\n",
    "    \n",
    "    return cvmodel, cvprediction,round(round(end-start)/60,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "88d5d696c93841c68e5627b53716643f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# check grid search results \n",
    "def cvmodel_result(cvmodel):\n",
    "    '''\n",
    "    INPUT:\n",
    "    Model - fitted model based on train data\n",
    "    \n",
    "    OUTPUT:\n",
    "    p_df - pandas data frame of grid search results\n",
    "    '''\n",
    "    f1 = cvmodel.avgMetrics\n",
    "    p_df = pd.DataFrame({p.name: v for p, v in param.items()} for param in cvmodel.getEstimatorParamMaps())\n",
    "    p_df['f1'] = f1\n",
    "    p_df = p_df.sort_values(by ='f1',ascending =False)\n",
    "    \n",
    "    return p_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9b72a07d77754d1b99bdb776ef148ec8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#feature importance\n",
    "def feature_importance(cvmodel,metric):\n",
    "    '''\n",
    "    INPUT:\n",
    "    Model - fitted model based on train data\n",
    "    \n",
    "    OUTPUT:\n",
    "    f_df - pandas data frame of feature metric (importance)\n",
    "    '''\n",
    "    f_df = pd.DataFrame()\n",
    "    feature = catcols + numcols\n",
    "    f_df['feature'] = feature\n",
    "    f_df['importance'] = metric \n",
    "    f_df = f_df.sort_values(by ='importance',ascending =False)\n",
    "\n",
    "    return f_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "76e888a4d46447bc8ad0a4bbde636069",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#plot confusion matrix\n",
    "def plot_confusion_matrix(cm, \n",
    "                          classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting 'normalize=True'.\n",
    "    INPUT:\n",
    "    cm - confusion matrix\n",
    "    classes - classes list\n",
    "    normalize - Normalization\n",
    "    title - plot title\n",
    "    cmap - plot color pattern\n",
    "    \n",
    "    OUTPUT:\n",
    "    print and plot the confusion matrix\n",
    "    \"\"\"\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    print(cm)\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], fmt),\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    #plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Logistic Regression Tuning\n",
    " - maxIter (maximum number of iterations, default=100): [10, 30]\n",
    " - threshold (Threshold in binary classification prediction, default = 0.5):[0.4, 0.5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "597ee6f1caad466abd68ec9017197cd5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.40813926905070297"
     ]
    }
   ],
   "source": [
    "# Extract the summary from the returned LogisticRegressionModel instance trained\n",
    "trainingSummary = model_lr.stages[-1].summary\n",
    "# check the model threshold to maximize F-Measure\n",
    "fMeasure = trainingSummary.fMeasureByThreshold\n",
    "maxFMeasure = fMeasure.groupBy().max('F-Measure').select('max(F-Measure)').head()\n",
    "bestThreshold = fMeasure.where(fMeasure['F-Measure'] == maxFMeasure['max(F-Measure)']) \\\n",
    "    .select('threshold').head()['threshold']\n",
    "bestThreshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "527fc4257696481ba98c995021ec6e8d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in thread cell_monitor-53:\n",
      "Traceback (most recent call last):\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 926, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 870, in run\n",
      "    self._target(*self._args, **self._kwargs)\n",
      "  File \"/opt/conda/lib/python3.7/site-packages/awseditorssparkmonitoringwidget-1.0-py3.7.egg/awseditorssparkmonitoringwidget/cellmonitor.py\", line 178, in cell_monitor\n",
      "    job_binned_stages[job_id][stage_id] = all_stages[stage_id]\n",
      "KeyError: 8864\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# parameter grid: maxIter,threshold.\n",
    "paramGrid_lr = ParamGridBuilder() \\\n",
    "    .addGrid(model_lr.stages[-1].maxIter,[10, 30]) \\\n",
    "    .addGrid(model_lr.stages[-1].threshold,[0.4,0.5]) \\\n",
    "    .build()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ba5fc88aa11749be9d6888d170c624f9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in thread cell_monitor-54:\n",
      "Traceback (most recent call last):\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 926, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 870, in run\n",
      "    self._target(*self._args, **self._kwargs)\n",
      "  File \"/opt/conda/lib/python3.7/site-packages/awseditorssparkmonitoringwidget-1.0-py3.7.egg/awseditorssparkmonitoringwidget/cellmonitor.py\", line 178, in cell_monitor\n",
      "    job_binned_stages[job_id][stage_id] = all_stages[stage_id]\n",
      "KeyError: 8897\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Logistic Regression Model Tuning time (mins):  12.95\n",
      "Summary Stats\n",
      "Accuracy = 0.9206088643486047\n",
      "Precision = 0.8367346938775511\n",
      "Recall = 0.817363344051447\n",
      "F1 Score = 0.8269355888093689"
     ]
    }
   ],
   "source": [
    "# tune the model \n",
    "cvmodel_lr, cvprediction_lr, cvtime_lr = model_tuning(pipeline_lr,paramGrid_lr,train, test)\n",
    "print('Logistic Regression Model Tuning time (mins): ',cvtime_lr)\n",
    "cvacc_lr, cvf1_lr = evaluation(cvprediction_lr)\n",
    "# F1 score on test data even decrease compared to the initial model. (overfitting on training data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e12f836b5ebe4f3bb83292b754630ad1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   maxIter  threshold        f1\n",
      "0       10        0.4  0.919546\n",
      "2       30        0.4  0.918898\n",
      "1       10        0.5  0.918354\n",
      "3       30        0.5  0.918292"
     ]
    }
   ],
   "source": [
    "# training model results\n",
    "cvmodel_result(cvmodel_lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6b17d9dc20b745deb0f137fecd19ec0a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in thread cell_monitor-56:\n",
      "Traceback (most recent call last):\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 926, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 870, in run\n",
      "    self._target(*self._args, **self._kwargs)\n",
      "  File \"/opt/conda/lib/python3.7/site-packages/awseditorssparkmonitoringwidget-1.0-py3.7.egg/awseditorssparkmonitoringwidget/cellmonitor.py\", line 178, in cell_monitor\n",
      "    job_binned_stages[job_id][stage_id] = all_stages[stage_id]\n",
      "KeyError: 10695\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Coefficients\n",
    "lr_imp = feature_importance(cvmodel_lr,cvmodel_lr.bestModel.stages[-1].coefficients)\n",
    "lr_imp['odds'] = np.exp(lr_imp['importance'])\n",
    "lr_imp.plot(kind ='barh',x ='feature',y='importance',color ='navy',legend=None,\\\n",
    "            title = 'Logistic Regression Coefficients (Full 12GB Dataset)',figsize = (8,5)).invert_yaxis();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#plot confusion_matrix class names\n",
    "class_names = cvprediction_gbt.select(\"label\").groupBy(\"label\")\\\n",
    "                        .count().sort('count', ascending=False).toPandas()\n",
    "class_names = class_names[\"label\"].values.tolist()\n",
    "class_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f76c82edd8524b35bb2d912e8d6a94bf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "array([[4898,  248],\n",
      "       [ 284, 1271]])"
     ]
    }
   ],
   "source": [
    "#plot confusion_matrix\n",
    "y_truelr = cvprediction_lr.select(\"label\").withColumn(\"label\",col(\"label\").cast(DoubleType()))\n",
    "y_truelr = y_truelr.toPandas()\n",
    "y_predlr = cvprediction_lr.select(\"prediction\")\n",
    "y_predlr = y_predlr.toPandas()\n",
    "\n",
    "cnf_matrix_lr = confusion_matrix(y_truelr, y_predlr)\n",
    "cnf_matrix_lr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bf0b8aa3ff0e4776808957b20110e068",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[4898  248]\n",
      " [ 284 1271]]"
     ]
    }
   ],
   "source": [
    "# Plot non-normalized confusion matrix\n",
    "plt.figure(figsize=(8,5))\n",
    "cm_lr = plot_confusion_matrix(cnf_matrix_lr, classes=class_names,\n",
    "                      title='Logistic Regression Confusion matrix\\nwithout normalization')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAyAAAAH0CAYAAADFQEl4AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAA9hAAAPYQGoP6dpAABe70lEQVR4nO3dd3hU1dbH8d8kIYVUaiK9KRCqBIWoNEEiTRAUC2pCUyR0qVelqryCSldQlCCCYAMpUnLpCqiUICLkSkchAYUQEiD1vH9wM5chCQQSznDk+/GZ57mzz5591pkMN1mz197HZhiGIQAAAAAwgYuzAwAAAABw5yABAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAAAAApiEBAQAAAGAaEhAAt1TTpk3VtGnTAhuvQoUKioiIKLDxINlsNo0ePdrZYTjFzz//rAceeEDe3t6y2WyKiYkp0PE3bNggm82mDRs2FOi4/wR38ucOuNORgAB3iKioKNlsNm3fvt3ZoVzXli1bNHr0aCUkJNzS81SoUEE2m83+8Pb21v33369PP/30lp4XjmJiYvTcc8+pbNmy8vDwUNGiRdWiRQvNmTNHGRkZt+y8aWlpevLJJ3XmzBlNmjRJ8+bNU/ny5W/Z+f6pzPr3CuCfw83ZAQD4Z1uzZs0Nv2bLli0aM2aMIiIiFBAQ4HAsNjZWLi4F991J3bp19corr0iSTp48qdmzZys8PFwpKSnq2bNngZ3ndnbx4kW5uTnn18Hs2bPVq1cvBQYG6vnnn9fdd9+t8+fPa+3aterevbtOnjypf/3rX7fk3AcPHtTRo0f10UcfqUePHrfkHI0bN9bFixfl7u5+S8a/HVzr3+u1OPNzB8C5+JcP4JYq6D+8PDw8CnS80qVL67nnnrM/j4iIUKVKlTRp0iTTE5Dk5GR5e3ubek5J8vT0NP2ckrRt2zb16tVLoaGh+u677+Tr62s/NmDAAG3fvl2//vrrLTv/qVOnJOmG/mi+US4uLk57f29HmZmZSk1NlaenJ+8LcAejBAuAg127dqlVq1by8/OTj4+Pmjdvrm3btmXr98svv6hJkyby8vJSmTJl9MYbb2jOnDmy2Ww6cuSIvV9Oa0CmTZumGjVqqHDhwipSpIjq16+vBQsWSJJGjx6tIUOGSJIqVqxoL4/KGjOnNSAJCQkaOHCgKlSoIA8PD5UpU0YvvPCC/vrrrxu+/hIlSqhatWo6ePCgQ3tmZqYmT56sGjVqyNPTU4GBgXrppZd09uzZbP1Gjx6tUqVKqXDhwmrWrJl+++23bHFnlcRt3LhRvXv3VsmSJVWmTBn78ZUrV6pRo0by9vaWr6+v2rRpo7179zqcKy4uTl27dlWZMmXk4eGhu+66S+3bt3d4/7dv366wsDAVL15cXl5eqlixorp16+YwTk61+Hn5HGRdww8//KBBgwapRIkS8vb21uOPP67Tp09f970eM2aMbDab5s+f75B8ZKlfv77De5acnKxXXnnFXqpVtWpVvfPOOzIMI9v19OnTR0uWLFHNmjXl4eGhGjVqaNWqVfY+ERERatKkiSTpySeflM1ms39Oc1u3FBERoQoVKji0LVy4UCEhIfL19ZWfn59q1aqlKVOm2I/ntgbkyy+/VEhIiLy8vFS8eHE999xz+vPPP7Odz8fHR3/++ac6dOggHx8flShRQoMHD85TaVqFChXUtm1bbdiwQfXr15eXl5dq1aplj+Wbb75RrVq15OnpqZCQEO3atcvh9b/88os9Iff09FRQUJC6deumv//+297nev9es34W8+fPV40aNeTh4WH/OVz5ubt48aKqVaumatWq6eLFi/bxz5w5o7vuuksPPPDALS3HA2AuZkAA2O3du1eNGjWSn5+fhg4dqkKFCmnWrFlq2rSpNm7cqAYNGkiS/vzzTzVr1kw2m00jRoyQt7e3Zs+enafZiY8++kj9+vXTE088of79++vSpUv65Zdf9OOPP+rZZ59Vx44d9Z///Eeff/65Jk2apOLFi0u6nBjkJCkpSY0aNdK+ffvUrVs31atXT3/99ZeWLl2qP/74w/76vEpPT9cff/yhIkWKOLS/9NJLioqKUteuXdWvXz8dPnxY06dP165du/TDDz+oUKFCkqQRI0ZowoQJateuncLCwrR7926FhYXp0qVLOZ6vd+/eKlGihEaOHKnk5GRJ0rx58xQeHq6wsDC9/fbbunDhgj744AM99NBD2rVrl/2P4E6dOmnv3r3q27evKlSooFOnTik6OlrHjh2zP2/ZsqVKlCih4cOHKyAgQEeOHNE333xzzfcgr5+DLH379lWRIkU0atQoHTlyRJMnT1afPn20aNGiXM9x4cIFrV27Vo0bN1a5cuWuGY8kGYahxx57TOvXr1f37t1Vt25drV69WkOGDNGff/6pSZMmOfT//vvv9c0336h3797y9fXV1KlT1alTJx07dkzFihXTSy+9pNKlS+utt95Sv379dN999ykwMPC6cVwpOjpazzzzjJo3b663335bkrRv3z798MMP6t+/f66vy/oc3XfffRo/frzi4+M1ZcoU/fDDD9q1a5fDjExGRobCwsLUoEEDvfPOO/r3v/+td999V5UrV9bLL7983RgPHDigZ599Vi+99JKee+45vfPOO2rXrp1mzpypf/3rX+rdu7ckafz48ercubNDiWN0dLQOHTqkrl27KigoSHv37tWHH36ovXv3atu2bbLZbHn697pu3Tp98cUX6tOnj4oXL54tiZMkLy8vzZ07Vw8++KBeffVVvffee5KkyMhInTt3TlFRUXJ1db3u9QKwCAPAHWHOnDmGJOPnn3/OtU+HDh0Md3d34+DBg/a2EydOGL6+vkbjxo3tbX379jVsNpuxa9cue9vff/9tFC1a1JBkHD582N7epEkTo0mTJvbn7du3N2rUqHHNWCdOnJhtnCzly5c3wsPD7c9HjhxpSDK++eabbH0zMzOveZ7y5csbLVu2NE6fPm2cPn3a2LNnj/H8888bkozIyEh7v82bNxuSjPnz5zu8ftWqVQ7tcXFxhpubm9GhQweHfqNHjzYkOcSd9fN46KGHjPT0dHv7+fPnjYCAAKNnz54OY8TFxRn+/v729rNnzxqSjIkTJ+Z6fYsXL77uz9wwDEOSMWrUKPvzvH4Osq6hRYsWDu/1wIEDDVdXVyMhISHXc+7evduQZPTv3/+asWVZsmSJIcl44403HNqfeOIJw2azGQcOHHC4Hnd3d4e2rPNNmzbN3rZ+/XpDkvHll186jHn1ZzZLeHi4Ub58efvz/v37G35+fg4/v6tlnWP9+vWGYRhGamqqUbJkSaNmzZrGxYsX7f2WL19uSDJGjhzpcD5JxtixYx3GvPfee42QkJBcz5mlfPnyhiRjy5Yt9rbVq1cbkgwvLy/j6NGj9vZZs2Y5xGkYhnHhwoVsY37++eeGJGPTpk32tmv9e5VkuLi4GHv37s3x2JWfO8MwjBEjRhguLi7Gpk2bjC+//NKQZEyePPm61wrAWijBAiDp8jeta9asUYcOHVSpUiV7+1133aVnn31W33//vRITEyVJq1atUmhoqOrWrWvvV7RoUXXp0uW65wkICNAff/yhn3/+uUDi/vrrr1WnTh09/vjj2Y7ZbLbrvn7NmjUqUaKESpQooVq1amnevHnq2rWrJk6caO/z5Zdfyt/fX4888oj++usv+yMkJEQ+Pj5av369JGnt2rVKT0+3f6ucpW/fvrmev2fPng7f7EZHRyshIUHPPPOMw7lcXV3VoEED+7m8vLzk7u6uDRs2ZCsDy5L1Tfry5cuVlpZ23fdCurHPQZYXX3zR4b1u1KiRMjIydPTo0VzPkzVGTqVXOfnuu+/k6uqqfv36ObS/8sorMgxDK1eudGhv0aKFKleubH9eu3Zt+fn56dChQ3k6X14EBAQoOTlZ0dHReX7N9u3bderUKfXu3dthDUSbNm1UrVo1rVixIttrevXq5fC8UaNGeb6O4OBghYaG2p9nzV49/PDDDjNPWe1Xjuvl5WX/35cuXdJff/2lhg0bSpJ27tyZp/NLUpMmTRQcHJynvqNHj1aNGjUUHh6u3r17q0mTJtl+5gCsjwQEgCTp9OnTunDhgqpWrZrtWPXq1ZWZmanjx49Lko4ePaoqVapk65dT29WGDRsmHx8f3X///br77rsVGRmpH3744abjPnjwoGrWrHnTr2/QoIGio6O1atUqvfPOOwoICNDZs2cdFs///vvvOnfunEqWLGlPVrIeSUlJ9sXMWX9wX/0+FC1aNFtJV5aKFSs6PP/9998lXf4D8epzrVmzxn4uDw8Pvf3221q5cqUCAwPVuHFjTZgwQXFxcfaxmjRpok6dOmnMmDEqXry42rdvrzlz5iglJSXX9+NGPgdZri6hyrrW3BIjSfLz85MknT9/Ptc+Vzp69KhKlSqVLWGpXr26/fi1YsqK61ox3ajevXvrnnvuUatWrVSmTBl169bNYZ1JTrLizOn9rVatWrbr8PT0zFZ+eCPXcfX74O/vL0kqW7Zsju1XjnvmzBn1799fgYGB8vLyUokSJeyf13PnzuXp/FL2z/i1uLu765NPPtHhw4d1/vx5+7oyAP8srAEBYKrq1asrNjZWy5cv16pVq/T111/r/fff18iRIzVmzBjT4ylevLhatGghSQoLC1O1atXUtm1bTZkyRYMGDZJ0eWF5yZIlNX/+/BzHyG19Sl5c+S1z1rmky+tAgoKCsvW/ctvSAQMGqF27dlqyZIlWr16t119/XePHj9e6det07733ymaz6auvvtK2bdu0bNkyrV69Wt26ddO7776rbdu2ycfH56bjvlJutfnGVYvDr1SlShW5ublpz549BRJDQcSUxWaz5djv6kXQJUuWVExMjFavXq2VK1dq5cqVmjNnjl544QXNnTv35gK/Sn7XPeT2+ry8P507d9aWLVs0ZMgQ1a1bVz4+PsrMzNSjjz5q/5zmxdWf8etZvXq1pMuzLr///vsNJTAArIEZEACSLv8RXbhwYcXGxmY7tn//frm4uNi/NS1fvrwOHDiQrV9ObTnx9vbWU089pTlz5ujYsWNq06aN3nzzTftC7Rv5xrNy5coFulVrmzZt1KRJE7311lv2ReGVK1fW33//rQcffFAtWrTI9qhTp44k2W9id/X78Pfff+f5G+ussqGSJUvmeK6rd2eqXLmyXnnlFa1Zs0a//vqrUlNT9e677zr0adiwod58801t375d8+fP1969e7Vw4cIcz38jn4P8KFy4sB5++GFt2rQp24xKTsqXL68TJ05kmzHZv3+//XhBKVKkSI431cuppMzd3V3t2rXT+++/r4MHD+qll17Sp59+muu/haw4c3p/Y2Njb5sbIZ49e1Zr167V8OHDNWbMGD3++ON65JFHHMryshTkDMUvv/yisWPHqmvXrrr33nvVo0ePG5ptAWANJCAAJF3+RrRly5b69ttvHbZxjY+P14IFC/TQQw/Zy2bCwsK0detWxcTE2PudOXMm1xmCK125had0+Q+44OBgGYZhX6eQdS+MvNxZuVOnTtq9e7cWL16c7Vhevu3OybBhw/T333/ro48+knT5m+CMjAyNGzcuW9/09HR7nM2bN5ebm5s++OADhz7Tp0/P87nDwsLk5+ent956K8d1G1nb2164cCHbzlqVK1eWr6+vvcTq7Nmz2d6DrHU7uZVh3cjnIL9GjRolwzD0/PPPKykpKdvxHTt22GcSWrdurYyMjGzv5aRJk2Sz2dSqVasCiUm6/D7u37/fYSvh3bt3ZysVvPqz7OLiotq1a0vK/f2tX7++SpYsqZkzZzr0Wblypfbt26c2bdoU1GXkS9YMydWfn8mTJ2freyP/Xq8lLS1NERERKlWqlKZMmaKoqCjFx8dr4MCB+RoXwO2HEizgDvPJJ5/kWKfev39/vfHGG4qOjtZDDz2k3r17y83NTbNmzVJKSoomTJhg7zt06FB99tlneuSRR9S3b1/7NrzlypXTmTNnrvmNaMuWLRUUFKQHH3xQgYGB2rdvn6ZPn642bdrY6/tDQkIkSa+++qqefvppFSpUSO3atcvxJn1DhgzRV199pSeffFLdunVTSEiIzpw5o6VLl2rmzJn22Ykb0apVK9WsWVPvvfeeIiMj1aRJE7300ksaP368YmJi1LJlSxUqVEi///67vvzyS02ZMkVPPPGEAgMD1b9/f7377rt67LHH9Oijj2r37t1auXKlihcvnqdviv38/PTBBx/o+eefV7169fT000+rRIkSOnbsmFasWKEHH3xQ06dP13/+8x81b95cnTt3VnBwsNzc3LR48WLFx8fr6aefliTNnTtX77//vh5//HFVrlxZ58+f10cffSQ/Pz+1bt061xjy+jnIrwceeEAzZsxQ7969Va1aNYc7oW/YsEFLly7VG2+8IUlq166dmjVrpldffVVHjhxRnTp1tGbNGn377bcaMGCAw4Lz/OrWrZvee+89hYWFqXv37jp16pRmzpypGjVqOCzA79Gjh86cOaOHH35YZcqU0dGjRzVt2jTVrVvXvjblaoUKFdLbb7+trl27qkmTJnrmmWfs2/BWqFDhtvlj28/Pz76uKC0tTaVLl9aaNWt0+PDhbH1v5N/rtbzxxhuKiYnR2rVr5evrq9q1a2vkyJF67bXX9MQTT1zzMwvAYpy1/RYAc2VtmZrb4/jx44ZhGMbOnTuNsLAww8fHxyhcuLDRrFkzh208s+zatcto1KiR4eHhYZQpU8YYP368MXXqVEOSERcXZ+939Zams2bNMho3bmwUK1bM8PDwMCpXrmwMGTLEOHfunMP448aNM0qXLm24uLg4bPF59Ta8hnF5C+A+ffoYpUuXNtzd3Y0yZcoY4eHhxl9//XXN96R8+fJGmzZtcjwWFRVlSDLmzJljb/vwww+NkJAQw8vLy/D19TVq1aplDB061Dhx4oS9T3p6uvH6668bQUFBhpeXl/Hwww8b+/btM4oVK2b06tUr288jty1y169fb4SFhRn+/v6Gp6enUblyZSMiIsLYvn27YRiG8ddffxmRkZFGtWrVDG9vb8Pf399o0KCB8cUXX9jH2Llzp/HMM88Y5cqVMzw8PIySJUsabdu2tY+RRTlsh5qXz0Fu13D11rPXs2PHDuPZZ581SpUqZRQqVMgoUqSI0bx5c2Pu3LlGRkaGvd/58+eNgQMH2vvdfffdxsSJE7Ntt6yrtlHOcvVnJ7dteA3DMD777DOjUqVKhru7u1G3bl1j9erV2bbh/eqrr4yWLVsaJUuWNNzd3Y1y5coZL730knHy5MnrvheLFi0y7r33XsPDw8MoWrSo0aVLF+OPP/5w6BMeHm54e3tni23UqFFGXn595/b5zun9OXz4cLZtnf/44w/j8ccfNwICAgx/f3/jySefNE6cOJHj5yW3f6+5/SyyjmWNs2PHDsPNzc3o27evQ5/09HTjvvvuM0qVKmWcPXv2utcMwBpshnGTNQoAcJUBAwZo1qxZSkpK4qZhV0hISFCRIkX0xhtv6NVXX3V2OAAAOBVrQADclIsXLzo8//vvvzVv3jw99NBDd3TycfX7Iv2vbv7qBeQAANyJWAMC4KaEhoaqadOmql69uuLj4/Xxxx8rMTFRr7/+urNDc6pFixYpKipKrVu3lo+Pj77//nt9/vnnatmypR588EFnhwcAgNORgAC4Ka1bt9ZXX32lDz/8UDabTfXq1dPHH3+sxo0bOzs0p6pdu7bc3Nw0YcIEJSYm2hemZy2mBgDgTscaEAAAAACmYQ0IAAAAANOQgAAAAAAwDQkIAOSiadOmed65qmnTpqpZs+atDQi5OnLkiGw2m6Kiouxto0ePztPNHwuas84LAFZBAgIAeXTixAmNHj1aMTExzg7FwXfffafRo0c7O4w7yoULFzR69Ght2LDB2aEAgOWQgABALtasWaM1a9bYn584cUJjxoy5LROQMWPGODuM285rr72W431ZCsKFCxc0ZsyYHBOQW3leAPgnYBteAMiFu7u7s0OwrEuXLsnd3V0uLs77nsvNzU1ubub/mnPWeQHAKpgBAfCP9ssvv8hms2np0qX2th07dtjvXXKlVq1aqUGDBvbnV64B2bBhg+677z5JUteuXWWz2bKtOZCk3377Tc2aNVPhwoVVunRpTZgwIVtMp06dUvfu3RUYGChPT0/VqVNHc+fOdeizYcMG2Wy2bN+wX73WISIiQjNmzJAke0zXW39QoUIFtW3bVt9//73uv/9+eXp6qlKlSvr000+z9T106JCefPJJFS1aVIULF1bDhg21YsWKHGNduHChXnvtNZUuXVqFCxdWYmKiIiIi5OPjo2PHjqlt27by8fFR6dKl7THv2bNHDz/8sLy9vVW+fHktWLDAYewzZ85o8ODBqlWrlnx8fOTn56dWrVpp9+7d17xGKftajIiICIf36MpHVglbamqqRo4cqZCQEPn7+8vb21uNGjXS+vXrHX4GJUqUkCSNGTMm2xg5rQFJT0/XuHHjVLlyZXl4eKhChQr617/+pZSUlJv+2QCAVZGAAPhHq1mzpgICArRp0yZ72+bNm+Xi4qLdu3crMTFRkpSZmaktW7bkeiPF6tWra+zYsZKkF198UfPmzdO8efMc+p89e1aPPvqo6tSpo3fffVfVqlXTsGHDtHLlSnufixcvqmnTppo3b566dOmiiRMnyt/fXxEREZoyZcoNX99LL72kRx55RJLsMc2bN++6rztw4ICeeOIJPfLII3r33XdVpEgRRUREaO/evfY+8fHxeuCBB7R69Wr17t1bb775pi5duqTHHntMixcvzjbmuHHjtGLFCg0ePFhvvfWWfQYpIyNDrVq1UtmyZTVhwgRVqFBBffr0UVRUlB599FHVr19fb7/9tnx9ffXCCy/o8OHD9jEPHTqkJUuWqG3btnrvvfc0ZMgQ7dmzR02aNNGJEydu+L268j3K+hlIUsmSJSVJiYmJmj17tpo2baq3335bo0eP1unTpxUWFmYvvStRooQ++OADSdLjjz9uH6tjx465nrtHjx4aOXKk6tWrp0mTJqlJkyYaP368nn766Zv62QCApRkA8A/Xpk0b4/7777c/79ixo9GxY0fD1dXVWLlypWEYhrFz505DkvHtt9/a+zVp0sRo0qSJ/fnPP/9sSDLmzJmT7RxNmjQxJBmffvqpvS0lJcUICgoyOnXqZG+bPHmyIcn47LPP7G2pqalGaGio4ePjYyQmJhqGYRjr1683JBnr1693OM/hw4ezxRAZGWncyP+dly9f3pBkbNq0yd526tQpw8PDw3jllVfsbQMGDDAkGZs3b7a3nT9/3qhYsaJRoUIFIyMjwyHWSpUqGRcuXHA4V3h4uCHJeOutt+xtZ8+eNby8vAybzWYsXLjQ3r5//35DkjFq1Ch726VLl+znufI98PDwMMaOHXvN92XUqFHXfF9+//13w9/f33jkkUeM9PR0wzAMIz093UhJSXHod/bsWSMwMNDo1q2bve306dPZYs3tvDExMYYko0ePHg79Bg8ebEgy1q1bZ2/L688GAKyMGRAA/3iNGjXSzp07lZycLEn6/vvv1bp1a9WtW1ebN2+WdHlWxGaz6aGHHrrp8/j4+Oi5556zP3d3d9f999+vQ4cO2du+++47BQUF6ZlnnrG3FSpUSP369VNSUpI2btx40+e/EcHBwWrUqJH9eYkSJVS1atVssd5///0O74mPj49efPFFHTlyRL/99pvDmOHh4fLy8srxfD169LD/74CAAFWtWlXe3t7q3Lmzvb1q1aoKCAhwiMHDw8O+jiQjI0N///23fHx8VLVqVe3cufMmr15KTk7W448/riJFiujzzz+Xq6urJMnV1dU+c5OZmakzZ84oPT1d9evXv+nzfffdd5KkQYMGObS/8sorkpStpC0vPxsAsDISEAD/eI0aNVJ6erq2bt2q2NhYnTp1So0aNVLjxo0dEpDg4GAVLVr0ps9TpkyZbLX/RYoU0dmzZ+3Pjx49qrvvvjvb4uzq1avbj5uhXLly2dpyirVq1arZ+uUWa8WKFXM8l6enp33NRBZ/f/8c3y9/f3+HGDIzMzVp0iTdfffd8vDwUPHixVWiRAn98ssvOnfu3HWuMnc9e/bUwYMHtXjxYhUrVszh2Ny5c1W7dm15enqqWLFiKlGihFasWHHT5zt69KhcXFxUpUoVh/agoCAFBARkex/z8rMBACsjAQHwj1e/fn15enpq06ZN2rx5s0qWLKl77rlHjRo10k8//aSUlBRt3rzZ4Vvnm5H1LfrVDMO44bFyW0iekZFxw2PlpCBjzZLb7Edu58pLDG+99ZYGDRqkxo0b67PPPtPq1asVHR2tGjVqKDMz86binDJlij7//HN99NFHqlu3rsOxzz77TBEREapcubI+/vhjrVq1StHR0Xr44Ydv+nxZ8npzwlvxswGA2wn7BAL4x8sqhdq8ebPKlStnTzQaNWqklJQUzZ8/X/Hx8bkuQM9SEHe3Ll++vH755RdlZmY6zILs37/ffly6/I23JCUkJDi8PqcZklt11+3y5csrNjY2W/vVsd5KX331lZo1a6aPP/7YoT0hIUHFixe/4fE2b96swYMHa8CAAfYF6Fefr1KlSvrmm28c3tdRo0Y59LuR97x8+fLKzMzU77//bp89ki4v8k9ISDDlfQSA2wkzIADuCI0aNdKPP/6o9evX2xOQ4sWLq3r16nr77bftfa7F29tbUvak4Ea0bt1acXFxWrRokb0tPT1d06ZNk4+Pj5o0aSLp8h+trq6uDrt3SdL7779/S+LKLdaffvpJW7dutbclJyfrww8/VIUKFRQcHFyg58uJq6trtm/+v/zyS/355583PNbJkyfVuXNnPfTQQ5o4cWKu55McZxt+/PFHh/dAkgoXLiwpb+9569atJUmTJ092aH/vvfckSW3atMlT/ADwT8EMCIA7QqNGjfTmm2/q+PHjDolG48aNNWvWLFWoUEFlypS55hiVK1dWQECAZs6cKV9fX3l7e6tBgwa5rn3IyYsvvqhZs2YpIiJCO3bsUIUKFfTVV1/phx9+0OTJk+Xr6yvp8lqIJ598UtOmTZPNZlPlypW1fPlynTp1KtuYISEhkqR+/fopLCxMrq6uOW7veqOGDx+uzz//XK1atVK/fv1UtGhRzZ07V4cPH9bXX39tyk0G27Ztq7Fjx6pr16564IEHtGfPHs2fP1+VKlW64bH69eun06dPa+jQoVq4cKHDsdq1a6t27dpq27atvvnmGz3++ONq06aNDh8+rJkzZyo4OFhJSUn2/l5eXgoODtaiRYt0zz33qGjRoqpZs6Zq1qyZ7bx16tRReHi4PvzwQyUkJKhJkyb66aefNHfuXHXo0EHNmjW78TcGACyMBATAHeGBBx6Qq6urChcurDp16tjbGzVqpFmzZuVp/UehQoU0d+5cjRgxQr169VJ6errmzJlzQwmIl5eXNmzYoOHDh2vu3LlKTExU1apVNWfOHEVERDj0nTZtmtLS0jRz5kx5eHioc+fOmjhxYrY/cjt27Ki+fftq4cKF+uyzz2QYRoEkIIGBgdqyZYuGDRumadOm6dKlS6pdu7aWLVtm2rf2//rXv5ScnKwFCxZo0aJFqlevnlasWKHhw4ff8FinT59WRkZGtt2opMslVrVr11ZERITi4uI0a9YsrV69WsHBwfrss8/05ZdfZrsp5OzZs9W3b18NHDhQqampGjVqVI4JSFbfSpUqKSoqSosXL1ZQUJBGjBiRrbQLAO4ENoNVbQAAAABMwhoQAAAAAKYhAQEAAABgGhIQAAAAAKYhAQEAAABgGhIQAAAAAKYhAQEAAABgGu4DAhSAzMxMnThxQr6+vrLZbM4OBwBgIsMwdP78eZUqVcqUG3Re7dKlS0pNTS2Qsdzd3eXp6VkgYwG5IQEBCsCJEydUtmxZZ4cBAHCi48ePq0yZMqae89KlS/LyLSalXyiQ8YKCgnT48GGSENxSJCBAAfD19ZUkuQeHy+bq7uRoAPMd2/COs0MAnOZ8YqKqVCxr/11gptTUVCn9gjyCw6X8/v7JSFXcb3OVmppKAoJbigQEKABZZVc2V3cSENyR/Pz8nB0C4HROLcF188z37x/DxtJgmINPGgAAAADTMAMCAABgdTZJ+Z2BYQ8VmIQEBAAAwOpsLpcf+R0DMAGfNAAAAACmYQYEAADA6my2AijBogYL5iABAQAAsDpKsGAhfNIAAAAAmIYZEAAAAKujBAsWQgICAABgeQVQgkVhDEzCJw0AAACAaZgBAQAAsDpKsGAhJCAAAABWxy5YsBA+aQAAAABMwwwIAACA1VGCBQshAQEAALA6SrBgIXzSAAAAAJiGGRAAAACrowQLFkICAgAAYHWUYMFC+KQBAAAAMA0zIAAAAFZnsxXADAglWDAHCQgAAIDVudguP/I7BmACSrAAAAAAmIYZEAAAAKtjEToshAQEAADA6tiGFxZCqgsAAADANMyAAAAAWB0lWLAQEhAAAACrowQLFkKqCwAAAMA0zIAAAABYHSVYsBASEAAAAKujBAsWQqoLAAAAwDTMgAAAAFgdJViwEBIQAAAAq6MECxZCqgsAAADANMyAAAAAWF4BlGDxvTRMQgICAABgdZRgwUJIdQEAAACYhhkQAAAAq7PZCmAXLGZAYA4SEAAAAKtjG15YCJ80AAAAAKZhBgQAAMDqWIQOCyEBAQAAsDpKsGAhfNIAAAAAmIYZEAAAAKujBAsWQgICAABgdZRgwUL4pAEAAAAwDTMgAAAAVkcJFiyEBAQAAMDibDabbCQgsAhKsAAAAACYhhkQAAAAi2MGBFZCAgIAAGB1tv8+8jsGYAJKsAAAAACYhhkQAAAAi6MEC1ZCAgIAAGBxJCCwEkqwAAAAAJiGGRAAAACLYwYEVkICAgAAYHEkILASSrAAAAAAmIYZEAAAAKvjPiCwEBIQAAAAi6MEC1ZCCRYAAAAA0zADAgAAYHE2mwpgBqRgYgGuhwQEAADA4mwqgBIsMhCYhBIsAAAAAKZhBgQAAMDiWIQOKyEBAQAAsDq24YWFUIIFAAAAwDQkIAAAAFb33xKs/DzyW4L1f//3f7LZbBowYIC97dKlS4qMjFSxYsXk4+OjTp06KT4+3uF1x44dU5s2bVS4cGGVLFlSQ4YMUXp6ukOfDRs2qF69evLw8FCVKlUUFRWVr1jhXCQgAAAAFpff5CO/a0h+/vlnzZo1S7Vr13ZoHzhwoJYtW6Yvv/xSGzdu1IkTJ9SxY0f78YyMDLVp00apqanasmWL5s6dq6ioKI0cOdLe5/Dhw2rTpo2aNWummJgYDRgwQD169NDq1atvOl44FwkIAAAAblpSUpK6dOmijz76SEWKFLG3nzt3Th9//LHee+89PfzwwwoJCdGcOXO0ZcsWbdu2TZK0Zs0a/fbbb/rss89Ut25dtWrVSuPGjdOMGTOUmpoqSZo5c6YqVqyod999V9WrV1efPn30xBNPaNKkSU65XuQfCQgAAIDFFeQMSGJiosMjJSXlmueOjIxUmzZt1KJFC4f2HTt2KC0tzaG9WrVqKleunLZu3SpJ2rp1q2rVqqXAwEB7n7CwMCUmJmrv3r32PlePHRYWZh8D1kMCAgAAYHW2AnpIKlu2rPz9/e2P8ePH53rahQsXaufOnTn2iYuLk7u7uwICAhzaAwMDFRcXZ+9zZfKRdTzr2LX6JCYm6uLFi9d6V3CbYhteAAAA2B0/flx+fn725x4eHrn269+/v6Kjo+Xp6WlWePgHYAYEAADA4gqyBMvPz8/hkVsCsmPHDp06dUr16tWTm5ub3NzctHHjRk2dOlVubm4KDAxUamqqEhISHF4XHx+voKAgSVJQUFC2XbGynl+vj5+fn7y8vPL93sF8JCAAAAAW54xdsJo3b649e/YoJibG/qhfv766dOli/9+FChXS2rVr7a+JjY3VsWPHFBoaKkkKDQ3Vnj17dOrUKXuf6Oho+fn5KTg42N7nyjGy+mSNAeuhBAsAAAA3zNfXVzVr1nRo8/b2VrFixezt3bt316BBg1S0aFH5+fmpb9++Cg0NVcOGDSVJLVu2VHBwsJ5//nlNmDBBcXFxeu211xQZGWmfeenVq5emT5+uoUOHqlu3blq3bp2++OILrVixwtwLRoEhAQEAALC4/N7HI2uMgjZp0iS5uLioU6dOSklJUVhYmN5//337cVdXVy1fvlwvv/yyQkND5e3trfDwcI0dO9bep2LFilqxYoUGDhyoKVOmqEyZMpo9e7bCwsIKPF6Yw2YYhuHsIACrS0xMlL+/vzxq9ZTN1d3Z4QCmO/vzdGeHADhNYmKiAov569y5cw6Lt806t7+/vwIj5snFvXC+xspMvaD4qOedch24s7AGBAAAAIBpKMECAACwuivu45GvMQATkIAAAABY3O26BgTICSVYAAAAAEzDDAgAAIDFMQMCK2EGBMBtZ3DXR3Rx13RNHNzJ3hZYzFcfj3tBh6Pf0l9b3tWWBcPUoXldh9fVrVZGyz/oo5ObJuiP9W9r+mvPyNvLcVeykOBy+m5mX53cNEEnNk7Q0hmRqnVPaTMuC7ghE98erwcb3qcSRXxVrlRJPdmpg/4TG5tjX8Mw1L5tK3kVsmnpt0scjm3/+We1atlcQcUDdFeJImrXOky/7N5twhXATM64ESFws0hAANxWQoLLqXunB/XLf/5waJ897gXdU6GknhwwS/WffEvfrovRZ293U52qZSRJd5Xw14qZfXXw+Gk1fv4dtY+coeDKQfpo7PP2Mby93PXtjEgdjzurxs+/o+Zd31PShUtaOiNSbm783yFuL5s3bVSvlyO18fttWr4yWulpaWrbuqWSk5Oz9Z02ZXKOfzwmJSWpfdtHVbZsOW364Uet3fC9fHx99VibMKWlpZlxGQCQDb9xAdw2vL3cNeetCPUe97kSEi86HGtYp5LeX7hR2/ce1ZE//9bbs1cr4fxF3RtcVpLUqlFNpaVnaMD4L/T70VPa8dsx9X1zkR5vca8qlS0uSapaMUjFArw17oPl+v3oKe07FKc3Z61UUHE/lburqOnXC1zL0hWr9Hx4hIJr1FDtOnX04cdROn7smHbt3OHQb3dMjKZMflczP/ok2xix+/frzJkzen30WN1TtaqCa9TQq6+NUnx8vI4dPWrWpcAMtgJ6ACYgAQFw25g84imt2vyr1v+Yvcxk2+5DeqJliIr4FZbNZtOTYSHy9HDTpu2/S5I83N2UlpahK++tejElVZL0QN3KkqT/HInXX2eTFN7hARVyc5WnRyFFdAjVvkMndfTEGROuELh5iefOSZKKFPlfsnzhwgVFvPCsJk+doaCgoGyvuadqVRUrVkxz53ys1NRUXbx4UVFzPla16tVVvkIFs0KHCSjBgpWQgAC4LTwZFqK61crq9WlLczz+3NBPVMjNVSc2TtC5Hydr2qtP66lBH+nQ8b8kSRt+ilVgMT8NfKG5Crm5KsDXS2/0ay9JCirhL0lKupCisJ5T9Ezr+3R22yT99cO7euSB6urQ531lZGSac6HATcjMzNSQVwYo9IEHVaNmTXv70FcGqmHDB9TusfY5vs7X11er/71Bny/4TEV8vVQ8wEfRa1ZpybKVcnNjHxoAzkECAsDpygQGaOKQTur6apRSUtNz7DMqsq0CfL3U6qWpevC5CZr62Tp9NqGbalQpJUnadyhOPUfOU7/nm+vM1vd05N9v6ciffyvur0QZmZeTC0+PQpo5qou27j6kJi+8o4e7vqffDp7UN1NflqdHIdOuF7hRA/pGau/eX/Xp/IX2tuXLlmrDhnWa+N7kXF938eJF9Xqxu0JDH9TG77dp3cYfFFyjpjq2b6OLFy/m+jpYDzMgsBK+/gD+a8aMGZo4caLi4uJUp04dTZs2Tffff7+zw7oj3Fu9nAKL+WnrgmH2Njc3Vz1Ur7J6PdVYtR8fp5efbqJ6nd7QvkNxkqQ9//lTD9arrJeeaqx+b17+o2zRqu1atGq7Shb1VfLFFBmG1O+5h3X4j78lSU+1qq9ypYqqSfi79lKt8BFROrlpgto1ra0vV+8QcLsZ0K+Pvvtuuf69bpPKlCljb9+wfp0OHTyooOIBDv2f6dxJDz7USGvWbtCizxfo2NEj2vj9Vrm4XP7Oce68BbqrRBEtW/qtOj/1tJmXglvIpgLYhpdFIDAJCQggadGiRRo0aJBmzpypBg0aaPLkyQoLC1NsbKxKlizp7PD+8db/FKuQJ950aPtwzHOKPRyvd6OiVdjz8la6mVes75CkjAxDLjn8wj115rwk6YX2DXUpNU1rt+2XJBX2dFdmpuGwTiTTMGQYynEcwJkMw9DA/n219NvFWvPvDapQsaLD8cFDh6trtx4ObfXvraUJ70xSm7btJF1eI+Li4uLwh2nW88xMyg4BOAclWICk9957Tz179lTXrl0VHBysmTNnqnDhwvrkk+y7yqDgJV1I0W8HTzo8ki+m6sy5ZP128KRij8TpwLFTmv7aM6pfo7wqlimu/s8/rOYNq2rZhv/dz6DXU41Vt1oZVSlXUi91bqxJwzpr5LSlOpd0udRk7bb9KuJXWJNHdFbVioGqXilIH45+TukZGdq4/T/OunwgRwP6Rmrhgs80d94C+fj6Ki4uTnFxcfbSqaCgINWoWdPhIUlly5WzJyvNWzyis2fPakDfSO3ft0+/7d2rF7t3lZubm5o0bea0a0PBowQLVsIMCO54qamp2rFjh0aMGGFvc3FxUYsWLbR169YcX5OSkqKUlBT788TExFse550sPT1THfp+oDf6tddXU16ST2EPHTx+Wj1GztPq73+z96tfs7xe69VGPoXdFXskXn3e/Fyfr/jZfvw/R+LVqf8svfpSK22Y+4oyMw3t3v+H2ke+r7i/+Bni9vLhrA8kSS2bN3Vsnz1Hz4dH5GmMqtWq6esly/TmuDFq2ihULi4uqlP3Xn27fJXuuuuuAo4YTlUQ2+iSf8AkJCC44/3111/KyMhQYGCgQ3tgYKD279+f42vGjx+vMWPGmBHeHSus5xSH5wePndYzg2df8zU9Xp933XHX/bhf637M+ecK3E4uphnX75SH1zRv8Yiat3ikIEICgAJBCRZwE0aMGKFz587ZH8ePH3d2SACAOxglWLASZkBwxytevLhcXV0VHx/v0B4fH5/jjb0kycPDQx4eHmaEBwDAdRVEAkECArMwA4I7nru7u0JCQrR27Vp7W2ZmptauXavQ0FAnRgYAAPDPwwwIIGnQoEEKDw9X/fr1df/992vy5MlKTk5W165dnR0aAADXZbNdfuR3DMAMJCCApKeeekqnT5/WyJEjFRcXp7p162rVqlXZFqYDAHA7upyA5LcEq4CCAa6DBAT4rz59+qhPnz7ODgMAAOAfjQQEAADA6gqgBIv7gMAsJCAAAAAWxy5YsBJ2wQIAAABgGmZAAAAALI5dsGAlJCAAAAAW5+Jik4tL/jIII5+vB/KKEiwAAAAApmEGBAAAwOIowYKVkIAAAABYHLtgwUoowQIAAABgGmZAAAAALI4SLFgJCQgAAIDFUYIFK6EECwAAAIBpmAEBAACwOGZAYCUkIAAAABbHGhBYCSVYAAAAAEzDDAgAAIDF2VQAJVhiCgTmIAEBAACwOEqwYCWUYAEAAAAwDTMgAAAAFscuWLASEhAAAACLowQLVkIJFgAAAADTMAMCAABgcZRgwUpIQAAAACyOEixYCSVYAAAAAEzDDAgAAIDFUYIFKyEBAQAAsLoCKMHiRugwCyVYAAAAAEzDDAgAAIDFUYIFKyEBAQAAsDh2wYKVUIIFAAAAwDTMgAAAAFgcJViwEhIQAAAAi6MEC1ZCCRYAAAAA0zADAgAAYHGUYMFKSEAAAAAsjgQEVkIJFgAAAADTMAMCAABgcSxCh5WQgAAAAFgcJViwEkqwAAAAAJiGGRAAAACLowQLVkICAgAAYHGUYMFKKMECAAAAYBpmQAAAACzOpgIowSqQSIDrIwEBAACwOBebTS75zEDy+3ogryjBAgAAAGAaZkAAAAAsjl2wYCUkIAAAABbHLliwEkqwAAAAAJiGGRAAAACLc7FdfuR3DMAMJCAAAABWZyuAEioSEJiEEiwAAADcsA8++EC1a9eWn5+f/Pz8FBoaqpUrV9qPX7p0SZGRkSpWrJh8fHzUqVMnxcfHO4xx7NgxtWnTRoULF1bJkiU1ZMgQpaenO/TZsGGD6tWrJw8PD1WpUkVRUVFmXB5uIRIQAAAAi8vaBSu/jxtRpkwZ/d///Z927Nih7du36+GHH1b79u21d+9eSdLAgQO1bNkyffnll9q4caNOnDihjh072l+fkZGhNm3aKDU1VVu2bNHcuXMVFRWlkSNH2vscPnxYbdq0UbNmzRQTE6MBAwaoR48eWr16dYG8b3AOm2EYhrODAKwuMTFR/v7+8qjVUzZXd2eHA5ju7M/TnR0C4DSJiYkKLOavc+fOyc/Pz/Rz+/v7q+WkdSrk5ZOvsdIuJmnNwIfzdR1FixbVxIkT9cQTT6hEiRJasGCBnnjiCUnS/v37Vb16dW3dulUNGzbUypUr1bZtW504cUKBgYGSpJkzZ2rYsGE6ffq03N3dNWzYMK1YsUK//vqr/RxPP/20EhIStGrVqnxdL5yHGRAAAADYJSYmOjxSUlKu+5qMjAwtXLhQycnJCg0N1Y4dO5SWlqYWLVrY+1SrVk3lypXT1q1bJUlbt25VrVq17MmHJIWFhSkxMdE+i7J161aHMbL6ZI0BayIBAQAAsLisXbDy+5CksmXLyt/f3/4YP358rufds2ePfHx85OHhoV69emnx4sUKDg5WXFyc3N3dFRAQ4NA/MDBQcXFxkqS4uDiH5CPreNaxa/VJTEzUxYsX8/OWwYnYBQsAAMDiCvJGhMePH3cowfLw8Mj1NVWrVlVMTIzOnTunr776SuHh4dq4cWO+4sA/HwkIAAAA7LJ2tcoLd3d3ValSRZIUEhKin3/+WVOmTNFTTz2l1NRUJSQkOMyCxMfHKygoSJIUFBSkn376yWG8rF2yruxz9c5Z8fHx8vPzk5eX101dH5yPBASWsnTp0jz3feyxx25hJAAA3D5uZhernMbIr8zMTKWkpCgkJESFChXS2rVr1alTJ0lSbGysjh07ptDQUElSaGio3nzzTZ06dUolS5aUJEVHR8vPz0/BwcH2Pt99953DOaKjo+1jwJpIQGApHTp0yFM/m82mjIyMWxsMAAC3CRebTS75zCBu9PUjRoxQq1atVK5cOZ0/f14LFizQhg0btHr1avn7+6t79+4aNGiQihYtKj8/P/Xt21ehoaFq2LChJKlly5YKDg7W888/rwkTJiguLk6vvfaaIiMj7WVfvXr10vTp0zV06FB169ZN69at0xdffKEVK1bk61rhXCQgsJTMzExnhwAAACSdOnVKL7zwgk6ePCl/f3/Vrl1bq1ev1iOPPCJJmjRpklxcXNSpUyelpKQoLCxM77//vv31rq6uWr58uV5++WWFhobK29tb4eHhGjt2rL1PxYoVtWLFCg0cOFBTpkxRmTJlNHv2bIWFhZl+vSg43AcE/wiXLl2Sp6en087PfUBwp+M+ILiT3Q73AWk3fUOB3AdkWZ+mTrkO3FnYhheWlZGRoXHjxql06dLy8fHRoUOHJEmvv/66Pv74YydHBwCAebJ2wcrvAzADCQgs680331RUVJQmTJggd/f/zTrUrFlTs2fPdmJkAAAAyA0JCCzr008/1YcffqguXbrI1dXV3l6nTh3t37/fiZEBAGCurF2w8vsAzMAidFjWn3/+ad97/EqZmZlKS0tzQkQAADiHM3bBAm4WMyCwrODgYG3evDlb+1dffaV7773XCREBAADgepgBgWWNHDlS4eHh+vPPP5WZmalvvvlGsbGx+vTTT7V8+XJnhwcAgGls/33kdwzADMyAwLLat2+vZcuW6d///re8vb01cuRI7du3T8uWLbPvQQ4AwJ2AXbBgJcyAwNIaNWqk6OhoZ4cBAACAPCIBgeVt375d+/btk3R5XUhISIiTIwIAwFwutsuP/I4BmIEEBJb1xx9/6JlnntEPP/yggIAASVJCQoIeeOABLVy4UGXKlHFugAAAmKQgSqgowYJZWAMCy+rRo4fS0tK0b98+nTlzRmfOnNG+ffuUmZmpHj16ODs8AAAA5IAZEFjWxo0btWXLFlWtWtXeVrVqVU2bNk2NGjVyYmQAAJiPCQxYBQkILKts2bI53nAwIyNDpUqVckJEAAA4ByVYsBJKsGBZEydOVN++fbV9+3Z72/bt29W/f3+98847TowMAAAAuWEGBJZSpEgRh29okpOT1aBBA7m5Xf4op6eny83NTd26dVOHDh2cFCUAAOZiFyxYCQkILGXy5MnODgEAgNsOJViwEhIQWEp4eLizQwAAAEA+kIDgH+HSpUtKTU11aPPz83NSNAAAmMv230d+xwDMQAICy0pOTtawYcP0xRdf6O+//852PCMjwwlRAQBgPhebTS75LKHK7+uBvGIXLFjW0KFDtW7dOn3wwQfy8PDQ7NmzNWbMGJUqVUqffvqps8MDAABADpgBgWUtW7ZMn376qZo2baquXbuqUaNGqlKlisqXL6/58+erS5cuzg4RAABT2Gz5vxEhEyAwCzMgsKwzZ86oUqVKki6v9zhz5owk6aGHHtKmTZucGRoAAKbK2gUrvw/ADCQgsKxKlSrp8OHDkqRq1arpiy++kHR5ZiQgIMCJkQEAACA3JCCwrK5du2r37t2SpOHDh2vGjBny9PTUwIEDNWTIECdHBwCAebJKsPL7AMzAGhBY1sCBA+3/u0WLFtq/f7927NihKlWqqHbt2k6MDAAAc7ELFqyEBAT/GOXLl1f58uWdHQYAAACugQQEljJ16tQ89+3Xr98tjAQAgNsHu2DBSkhAYCmTJk3KUz+bzUYCAgC4YxTELlbsggWzkIDAUrJ2vbpdHV0/UX5+fs4OAzDdr8fPOTsEwGmSzic6OwTAUkhAAAAALM5F+d/alK1RYRY+awAAAABMwwwIAACAxbEGBFZCAgIAAGBxNpvkwi5YsAhKsAAAAACYhgQElrZ582Y999xzCg0N1Z9//ilJmjdvnr7//nsnRwYAgHlcbAXzAMxAAgLL+vrrrxUWFiYvLy/t2rVLKSkpkqRz587prbfecnJ0AACYJ2sNSH4fgBlIQGBZb7zxhmbOnKmPPvpIhQoVsrc/+OCD2rlzpxMjAwAAQG5YhA7Lio2NVePGjbO1+/v7KyEhwfyAAABwkoIooaIEC2ZhBgSWFRQUpAMHDmRr//7771WpUiUnRAQAgHPYbAXzAMxAAgLL6tmzp/r3768ff/xRNptNJ06c0Pz58zV48GC9/PLLzg4PAAAAOaAEC5Y1fPhwZWZmqnnz5rpw4YIaN24sDw8PDR48WH379nV2eAAAmMbFZpNLPqcw8vt6IK9IQGBZNptNr776qoYMGaIDBw4oKSlJwcHB8vHxcXZoAACYykX5L2uhLAZmIQGB5bm7uys4ONjZYQAAACAPSEBgWc2aNbvmnuXr1q0zMRoAAJynIBaRU4EFs5CAwLLq1q3r8DwtLU0xMTH69ddfFR4e7pygAABwAhcVwBoQkYHAHCQgsKxJkybl2D569GglJSWZHA0AAADygvVG+Md57rnn9Mknnzg7DAAATMN9QGAlzIDgH2fr1q3y9PR0dhgAAJiGO6HDSkhAYFkdO3Z0eG4Yhk6ePKnt27fr9ddfd1JUAAAAuBYSEFiWv7+/w3MXFxdVrVpVY8eOVcuWLZ0UFQAA5rPZ8n8jQUqwYBYSEFhSRkaGunbtqlq1aqlIkSLODgcAAKdiG15YCYvQYUmurq5q2bKlEhISnB0KAAAAbgAJCCyrZs2aOnTokLPDAADA6bIWoef3AZiBBASW9cYbb2jw4MFavny5Tp48qcTERIcHAAB3ClsB/QeYgTUgsJyxY8fqlVdeUevWrSVJjz32mGxXFK4ahiGbzaaMjAxnhQgAAIBckIDAcsaMGaNevXpp/fr1zg4FAIDbAvcBgZWQgMByDMOQJDVp0sTJkQAAcHsgAYGVsAYElmRjr0AAAABLYgYElnTPPfdcNwk5c+aMSdEAAOBcNpst31/O8eUezEICAksaM2ZMtjuhAwBwp6IEC1ZCAgJLevrpp1WyZElnhwEAAIAbRAICy2GKGAAARzbb5Ud+xwDMQAICy8naBQsAAFzmYrPJJZ8ZRH5fD+QVCQgsJzMz09khAAAA4CaRgAAAAFgci9BhJSQgAAAAVlcAa0BEAgKTcCNCAAAAAKZhBgQAAMDiXGSTSz6nMPL7eiCvSEAAAAAsjm14YSWUYAEAAAAwDTMgAAAAFscuWLASZkAAAAAsLutGhPl93Ijx48frvvvuk6+vr0qWLKkOHTooNjbWoc+lS5cUGRmpYsWKycfHR506dVJ8fLxDn2PHjqlNmzYqXLiwSpYsqSFDhig9Pd2hz4YNG1SvXj15eHioSpUqioqKuqn3CbcHEhAAAADcsI0bNyoyMlLbtm1TdHS00tLS1LJlSyUnJ9v7DBw4UMuWLdOXX36pjRs36sSJE+rYsaP9eEZGhtq0aaPU1FRt2bJFc+fOVVRUlEaOHGnvc/jwYbVp00bNmjVTTEyMBgwYoB49emj16tWmXi8Kjs0wDMPZQQBWl5iYKH9/f8X9lSA/Pz9nhwOYbu8fic4OAXCapPOJalannM6dO2f674Cs3z9T1u6Rl7dvvsa6mHxe/ZvXuunrOH36tEqWLKmNGzeqcePGOnfunEqUKKEFCxboiSeekCTt379f1atX19atW9WwYUOtXLlSbdu21YkTJxQYGChJmjlzpoYNG6bTp0/L3d1dw4YN04oVK/Trr7/az/X0008rISFBq1atytc1wzmYAQEAALA4FxVACVY+t+E9d+6cJKlo0aKSpB07digtLU0tWrSw96lWrZrKlSunrVu3SpK2bt2qWrVq2ZMPSQoLC1NiYqL27t1r73PlGFl9ssaA9bAIHQAAAHaJiY4zmh4eHvLw8LjmazIzMzVgwAA9+OCDqlmzpiQpLi5O7u7uCggIcOgbGBiouLg4e58rk4+s41nHrtUnMTFRFy9elJeX141dIJyOGRAAAACLy7oPSH4fklS2bFn5+/vbH+PHj7/u+SMjI/Xrr79q4cKFt/hK8U/ADAgAAIDFuSj/3ypnvf748eMOa0CuN/vRp08fLV++XJs2bVKZMmXs7UFBQUpNTVVCQoLDLEh8fLyCgoLsfX766SeH8bJ2ybqyz9U7Z8XHx8vPz4/ZD4tiBgQAAAB2fn5+Do/cEhDDMNSnTx8tXrxY69atU8WKFR2Oh4SEqFChQlq7dq29LTY2VseOHVNoaKgkKTQ0VHv27NGpU6fsfaKjo+Xn56fg4GB7nyvHyOqTNQashxkQAAAAi7PZbLLd4H08chrjRkRGRmrBggX69ttv5evra1+z4e/vLy8vL/n7+6t79+4aNGiQihYtKj8/P/Xt21ehoaFq2LChJKlly5YKDg7W888/rwkTJiguLk6vvfaaIiMj7YlPr169NH36dA0dOlTdunXTunXr9MUXX2jFihX5ul44DwkIAACAxdn++8jvGDfigw8+kCQ1bdrUoX3OnDmKiIiQJE2aNEkuLi7q1KmTUlJSFBYWpvfff9/e19XVVcuXL9fLL7+s0NBQeXt7Kzw8XGPHjrX3qVixolasWKGBAwdqypQpKlOmjGbPnq2wsLCbuUzcBrgPCFAAuA8I7nTcBwR3stvhPiAz1++Vl08+7wOSdF69mtVwynXgzsIMCAAAgMVl3csjv2MAZiABAQAA+AcgfYBVsAsWAAAAANMwAwIAAGBxV95IMD9jAGYgAQEAALA4Z2zDC9wsSrAAAAAAmIYZEAAAAItzUf6/VeZbaZiFBAQAAMDiKMGClZDsAgAAADANMyAAAAAWZ1P+7wPC/AfMQgICAABgcZRgwUoowQIAAABgGmZAAAAALI5dsGAlJCAAAAAWRwkWrIRkFwAAAIBpmAEBAACwOHbBgpWQgAAAAFiczXb5kd8xADNQggUAAADANMyAAAAAWJyLbHLJZxFVfl8P5BUJCAAAgMVRggUroQQLAAAAgGmYAQEAALA423//y+8YgBlIQAAAACyOEixYCSVYAAAAAEzDDAgAAIDF2QpgFyxKsGAWEhAAAACLowQLVkIJFgAAAADTMAMCAABgccyAwEpIQAAAACyObXhhJZRgAQAAADANMyAAAAAW52K7/MjvGIAZSEAAAAAsjhIsWAklWAAAAABMQwIC4LY08e3xeij0fpUs6qfypQPVudPj+k9srEOfuLg4dY94QRXK3qXiAT4KvT9ES775OsfxUlJS1KD+vSrs7qLdMTEmXAFwY3b+9IMG9nhKrRpW032VArRhzXL7sfS0NE37v1F6+tEH1KhGKbVqWE2jXnlJp+NP2vvs2LZZ91UKyPGxd/dOSVJKyiWNHvKynn70ATW8u5gGv/Ss6deJWyNrF6z8PgAzkIAAuC1t3rxJL73cWxs2b9Wy79YoLT1N7dqEKTk52d6nZ7dw/ec/sfrym2/1885f1L7D43ru2acUs2tXtvFeHTFUd5UqZeYlADfk4oULuqd6LQ0dMzHbsUsXL2j/3t3q3neI5i3bqAkfzNPRQwf0Ss9n7H1q12uglT/GOjzaP/WCSpUtr+Da90qSMjMy5OnhpaciXtJ9DzY169JgApv+V4Z18/8B5mANCIDb0tLlKx2efzh7jsqXDtSunTv0UKPGkqRtW7doyrT3dd9990uShv/rNU2fOlm7du1Q3Xvvtb929aqVWhsdrQWLvtKaVY7jAreLB5s+ogebPpLjMR8/f82Yt8ShbcjoiYp4/GHF/XlcQaXLqpC7u4qXCLQfT09L06Z/f6fOL7wo23+/2vYq7K3hb7wnSdq9Y5uSEs/dmosBgGtgBgSAJSSeu/yHUpEiRe1tDUMf0FdffaEzZ84oMzNTXy5aqEuXLqlx46b2PvHx8Yp8+UXNjvpUhQsXNjts4JZJOp8om80mHz//HI9v+vd3Onf2jNo90cXkyOAMWbtg5fcBmIEEBMBtLzMzU0MGD1ToAw+qRs2a9vZ5CxYpPS1NZYKKK8DHU30je2nhl9+ocpUqkiTDMPRij67q0fMlhYTUd1b4QIFLSbmk6RNGqWW7J+Tj65djn2+/+EwNGzVX4F2lTY4OzpD/8iuKsGAeEhBA0qZNm9SuXTuVKlVKNptNS5YscXZIuMKAfpH6be+vmvvZ5w7tY0e/roSEBK1YFa3vt/6svv0H6vlnn9Kve/ZIkj6YMU1J589ryLARzggbuCXS09I0ok+EDMPQ8HHv5tgn/uSf2rZ5rdp3fs7k6ADg+khAAEnJycmqU6eOZsyY4exQcJWB/fto5XcrtGrNOpUpU8befujgQc18f4Zmfvixmj3cXLXr1NGrr49SvZD6mjXz8s9xw/r1+nHbVgX4eMrXq5BqVr9bkvRQ6H3q2S3CGZcD5Et6WppG9I1Q3J/HNf3TJbnOfiz7ar78ixRV4xatTY4QzsIuWLASFqEDklq1aqVWrVo5OwxcwTAMDRrQV0u/XaLV0etVoWJFh+MXLlyQJLm4OH6P4urqqszMTEnSu5OmaNSYcfZjJ0+e0GNtHtW8+QtV//4Gt/gKgIKVlXwcO3JIM+cvU8AV66GuZBiGln01X60ff1puhQqZHCWcxfbfR37HAMxAAgLgtjSgX6S+WPi5vvh6iXx8fRUXFydJ8vf3l5eXl6pWq6bKVaqob2QvvfX2RBUrWkzLli7R2n9H6+slyyRJZcuVcxjTx8dHklSxUmWH2RTgdnAhOUnHjx6yPz9x/Khif/tF/v5FVLxkkIZFvqD9e3/RpNkLlZGZob9Ox0uS/P2LqJC7u/11P2/ZpBPHj6rDUy/keJ5Dv+9XWlqqEhPO6kJykmJ/+0WSVDW49i28OgD4HxIQ4CakpKQoJSXF/jwxMdGJ0fwzfTRrpiQprEUzh/ZZsz/R8y9EqFChQlr87Qq9/uoIPfn4Y0pKSlLlylX00cdRerQVZSewnn17dqnXs+3szye9+aokqU2nZ/Ri/+Ha9O/LW0h3adPI4XUzFyxTSMP/tS39Yp5qhzRQhcr35HieAd2e1Mk/j9ufP9f28rbWPx9KKJDrgHO4yCaXfNZQuTAHApOQgAA3Yfz48RozZoyzw/hHu5Caed0+Ve6+W59/8VWexyxfoUKexgWcIaRho2smAXlNEN6YMvuax5du3nMDUcEqKMGClbAIHbgJI0aM0Llz5+yP48ePX/9FAAAAYAYEuBkeHh7y8PBwdhgAAFzGFAgshAQEkJSUlKQDBw7Ynx8+fFgxMTEqWrSoyl21kBkAgNtNQdxIkBsRwiwkIICk7du3q1mz/y12HjRokCQpPDxcUVFRTooKAADgn4cEBJDUtGlTGYbh7DAAALg5BXEjQSZAYBISEAAAAItjCQishF2wAAAAAJiGGRAAAACrYwoEFkICAgAAYHHsggUroQQLAAAAgGmYAQEAALA4WwHsgpXvXbSAPCIBAQAAsDiWgMBKKMECAAAAYBpmQAAAAKyOKRBYCAkIAACAxbELFqyEEiwAAAAApmEGBAAAwOLYBQtWQgICAABgcSwBgZVQggUAAADANMyAAAAAWB1TILAQEhAAAACLYxcsWAklWAAAAABMwwwIAACAxbELFqyEBAQAAMDiWAICK6EECwAAAIBpmAEBAACwOqZAYCEkIAAAABbHLliwEkqwAAAAAJiGBAQAAMDisnbByu/jRm3atEnt2rVTqVKlZLPZtGTJEofjhmFo5MiRuuuuu+Tl5aUWLVro999/d+hz5swZdenSRX5+fgoICFD37t2VlJTk0OeXX35Ro0aN5OnpqbJly2rChAk3HixuGyQgAAAAFmcroMeNSk5OVp06dTRjxowcj0+YMEFTp07VzJkz9eOPP8rb21thYWG6dOmSvU+XLl20d+9eRUdHa/ny5dq0aZNefPFF+/HExES1bNlS5cuX144dOzRx4kSNHj1aH3744U1EjNsBa0AAAABwU1q1aqVWrVrleMwwDE2ePFmvvfaa2rdvL0n69NNPFRgYqCVLlujpp5/Wvn37tGrVKv3888+qX7++JGnatGlq3bq13nnnHZUqVUrz589XamqqPvnkE7m7u6tGjRqKiYnRe++955CowDqYAQEAALA6Z02BXMPhw4cVFxenFi1a2Nv8/f3VoEEDbd26VZK0detWBQQE2JMPSWrRooVcXFz0448/2vs0btxY7u7u9j5hYWGKjY3V2bNnCzZomIIZEAAAAIsryF2wEhMTHdo9PDzk4eFxw+PFxcVJkgIDAx3aAwMD7cfi4uJUsmRJh+Nubm4qWrSoQ5+KFStmGyPrWJEiRW44NjgXMyAAAACwK1u2rPz9/e2P8ePHOzsk/MMwAwIAAGBxN7uL1dVjSNLx48fl5+dnb7+Z2Q9JCgoKkiTFx8frrrvusrfHx8erbt269j6nTp1yeF16errOnDljf31QUJDi4+Md+mQ9z+oDa2EGBAAAwOIKcgmIn5+fw+NmE5CKFSsqKChIa9eutbclJibqxx9/VGhoqCQpNDRUCQkJ2rFjh73PunXrlJmZqQYNGtj7bNq0SWlpafY+0dHRqlq1KuVXFkUCAgAAgJuSlJSkmJgYxcTESLq88DwmJkbHjh2TzWbTgAED9MYbb2jp0qXas2ePXnjhBZUqVUodOnSQJFWvXl2PPvqoevbsqZ9++kk//PCD+vTpo6efflqlSpWSJD377LNyd3dX9+7dtXfvXi1atEhTpkzRoEGDnHTVyC9KsAAAAKyuIHaxuonXb9++Xc2aNbM/z0oKwsPDFRUVpaFDhyo5OVkvvviiEhIS9NBDD2nVqlXy9PS0v2b+/Pnq06ePmjdvLhcXF3Xq1ElTp061H/f399eaNWsUGRmpkJAQFS9eXCNHjmQLXguzGYZhODsIwOoSExPl7++vuL8SHOpmgTvF3j8Sr98J+IdKOp+oZnXK6dy5c6b/Dsj6/bPz9zj5+Obv3EnnE1Xv7iCnXAfuLJRgAQAAADANJVgAAABWVwC7YBX0jQiB3JCAAAAAWJyTloAAN4USLAAAAACmYQYEAADA6pgCgYWQgAAAAFic7b//5XcMwAyUYAEAAAAwDTMgAAAAFmcrgF2w8r2LFpBHJCAAAAAWxxIQWAklWAAAAABMwwwIAACA1TEFAgshAQEAALA4dsGClVCCBQAAAMA0zIAAAABYnE0FsAtWgUQCXB8JCAAAgMWxBARWQgkWAAAAANMwAwIAAGBx3IgQVkICAgAAYHkUYcE6KMECAAAAYBpmQAAAACyOEixYCQkIAACAxVGABSuhBAsAAACAaZgBAQAAsDhKsGAlJCAAAAAWZ/vvf/kdAzADJVgAAAAATMMMCAAAgNWxCh0WQgICAABgceQfsBJKsAAAAACYhhkQAAAAi2MXLFgJCQgAAIDFsQsWrIQSLAAAAACmYQYEAADA6liFDgshAQEAALA48g9YCSVYAAAAAEzDDAgAAIDFsQsWrIQEBAAAwPLyvwsWRVgwCyVYAAAAAEzDDAgAAIDFUYIFK2EGBAAAAIBpSEAAAAAAmIYSLAAAAIujBAtWQgICAABgcbYC2AUr/7toAXlDCRYAAAAA0zADAgAAYHGUYMFKSEAAAAAszqb830aQ/ANmoQQLAAAAgGmYAQEAALA6pkBgISQgAAAAFscuWLASSrAAAAAAmIYZEAAAAItjFyxYCQkIAACAxbEEBFZCCRYAAAAA0zADAgAAYHVMgcBCSEAAAAAsjl2wYCWUYAEAAAAwDTMgQAEwDEOSdP58opMjAZwjic8+7mDJSecl/e93gTOcP5+Y712s+B0Gs5CAAAXg/PnLv3zurljOyZEAAJzl/Pnz8vf3N/Wc7u7uCgoK0t0VyxbIeEFBQXJ3dy+QsYDc2AxnpuvAP0RmZqZOnDghX19f2dhI3XSJiYkqW7asjh8/Lj8/P2eHA5iOfwPOZRiGzp8/r1KlSsnFxfzq9kuXLik1NbVAxnJ3d5enp2eBjAXkhhkQoAC4uLioTJkyzg7jjufn58cfX7ij8W/Aecye+biSp6cnSQMshUXoAAAAAExDAgIAAADANCQgACzPw8NDo0aNkoeHh7NDAZyCfwMArIRF6AAAAABMwwwIAAAAANOQgAAAAAAwDQkIAAAAANOQgAAAAAAwDQkIAMubMWOGKlSoIE9PTzVo0EA//fSTs0MCTLFp0ya1a9dOpUqVks1m05IlS5wdEgBcFwkIAEtbtGiRBg0apFGjRmnnzp2qU6eOwsLCdOrUKWeHBtxyycnJqlOnjmbMmOHsUAAgz9iGF4ClNWjQQPfdd5+mT58uScrMzFTZsmXVt29fDR8+3MnRAeax2WxavHixOnTo4OxQAOCamAEBYFmpqanasWOHWrRoYW9zcXFRixYttHXrVidGBgAAckMCAsCy/vrrL2VkZCgwMNChPTAwUHFxcU6KCgAAXAsJCAAAAADTkIAAsKzixYvL1dVV8fHxDu3x8fEKCgpyUlQAAOBaSEAAWJa7u7tCQkK0du1ae1tmZqbWrl2r0NBQJ0YGAABy4+bsAAAgPwYNGqTw8HDVr19f999/vyZPnqzk5GR17drV2aEBt1xSUpIOHDhgf3748GHFxMSoaNGiKleunBMjA4DcsQ0vAMubPn26Jk6cqLi4ONWtW1dTp05VgwYNnB0WcMtt2LBBzZo1y9YeHh6uqKgo8wMCgDwgAQEAAABgGtaAAAAAADANCQgAAAAA05CAAAAAADANCQgAAAAA05CAAAAAADANCQgAAAAA05CAAAAAADANCQgA3CEiIiLUoUMH+/OmTZtqwIABpsexYcMG2Ww2JSQk5NrHZrNpyZIleR5z9OjRqlu3br7iOnLkiGw2m2JiYvI1DgDg2khAAMCJIiIiZLPZZLPZ5O7uripVqmjs2LFKT0+/5ef+5ptvNG7cuDz1zUvSAABAXrg5OwAAuNM9+uijmjNnjlJSUvTdd98pMjJShQoV0ogRI7L1TU1Nlbu7e4Gct2jRogUyDgAAN4IZEABwMg8PDwUFBal8+fJ6+eWX1aJFCy1dulTS/8qm3nzzTZUqVUpVq1aVJB0/flydO3dWQECAihYtqvbt2+vIkSP2MTMyMjRo0CAFBASoWLFiGjp0qAzDcDjv1SVYKSkpGjZsmMqWLSsPDw9VqVJFH3/8sY4cOaJmzZpJkooUKSKbzaaIiAhJUmZmpsaPH6+KFSvKy8tLderU0VdffeVwnu+++0733HOPvLy81KxZM4c482rYsGG65557VLhwYVWqVEmvv/660tLSsvWbNWuWypYtq8KFC6tz5846d+6cw/HZs2erevXq8vT0VLVq1fT+++/fcCwAgPwhAQGA24yXl5dSU1Ptz9euXavY2FhFR0dr+fLlSktLU1hYmHx9fbV582b98MMP8vHx0aOPPmp/3bvvvquoqCh98skn+v7773XmzBktXrz4mud94YUX9Pnnn2vq1Knat2+fZs2aJR8fH5UtW1Zff/21JCk2NlYnT57UlClTJEnjx4/Xp59+qpkzZ2rv3r0aOHCgnnvuOW3cuFHS5USpY8eOateunWJiYtSjRw8NHz78ht8TX19fRUVF6bffftOUKVP00UcfadKkSQ59Dhw4oC+++ELLli3TqlWrtGvXLvXu3dt+fP78+Ro5cqTefPNN7du3T2+99ZZef/11zZ0794bjAQDkgwEAcJrw8HCjffv2hmEYRmZmphEdHW14eHgYgwcPth8PDAw0UlJS7K+ZN2+eUbVqVSMzM9PelpKSYnh5eRmrV682DMMw7rrrLmPChAn242lpaUaZMmXs5zIMw2jSpInRv39/wzAMIzY21pBkREdH5xjn+vXrDUnG2bNn7W2XLl0yChcubGzZssWhb/fu3Y1nnnnGMAzDGDFihBEcHOxwfNiwYdnGupokY/HixbkenzhxohESEmJ/PmrUKMPV1dX4448/7G0rV640XFxcjJMnTxqGYRiVK1c2FixY4DDOuHHjjNDQUMMwDOPw4cOGJGPXrl25nhcAkH+sAQEAJ1u+fLl8fHyUlpamzMxMPfvssxo9erT9eK1atRzWfezevVsHDhyQr6+vwziXLl3SwYMHde7cOZ08eVINGjSwH3Nzc1P9+vWzlWFliYmJkaurq5o0aZLnuA8cOKALFy7okUcecWhPTU3VvffeK0nat2+fQxySFBoamudzZFm0aJGmTp2qgwcPKikpSenp6fLz83PoU65cOZUuXdrhPJmZmYqNjZWvr68OHjyo7t27q2fPnvY+6enp8vf3v+F4AAA3jwQEAJysWbNm+uCDD+Tu7q5SpUrJzc3x/5q9vb0dniclJSkkJETz58/PNlaJEiVuKgYvL68bfk1SUpIkacWKFQ5/+EuX17UUlK1bt6pLly4aM2aMwsLC5O/vr4ULF+rdd9+94Vg/+uijbAmRq6trgcUKALg+EhAAcDJvb29VqVIlz/3r1aunRYsWqWTJktlmAbLcdddd+vHHH9W4cWNJl7/p37Fjh+rVq5dj/1q1aikzM1MbN25UixYtsh3PmoHJyMiwtwUHB8vDw0PHjh3LdeakevXq9gX1WbZt23b9i7zCli1bVL58eb366qv2tqNHj2brd+zYMZ04cUKlSpWyn8fFxUVVq1ZVYGCgSpUqpUOHDqlLly43dH4AQMFiEToAWEyXLl1UvHhxtW/fXps3b9bhw4e1YcMG9evXT3/88YckqX///vq///s/LVmyRPv371fv3r2veQ+PChUqKDw8XN26ddOSJUvsY37xxReSpPLly8tms2n58uU6ffq0kpKS5Ovrq8GDB2vgwIGaO3euDh48qJ07d2ratGn2hd29evXS77//riFDhig2NlYLFixQVFTUDV3v3XffrWPHjmnhwoU6ePCgpk6dmuOCek9PT4WHh2v37t3avHmz+vXrp86dOysoKEiSNGbMGI0fP15Tp07Vf/7zH+3Zs0dz5szRe++9d0PxAADyhwQEACymcOHC2rRpk8qVK6eOHTuqevXq6t69uy5dumSfEXnllVf0/PPPKzw8XKGhofL19dXjjz9+zXE/+OADPfHEE+rdu7eqVaumnj17Kjk5WZJUunRpjRkzRsOHD1dgYKD69OkjSRo3bpxef/11jR8/XtWrV9ejjz6qFStWqGLFipIur8v4+uuvtWTJEtWpU0czZ87UW2+9dUPX+9hjj2ngwIHq06eP6tatqy1btuj111/P1q9KlSrq2LGjWrdurZYtW6p27doO2+z26NFDs2fP1pw5c1SrVi01adJEUVFR9lgBAOawGbmtSAQAAACAAsYMCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMA0JCAAAAADTkIAAAAAAMM3/A+qk5K4B0hJ0AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "99442044113048c5a68a6806e50e279c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized confusion matrix\n",
      "[[0.95180723 0.04819277]\n",
      " [0.18263666 0.81736334]]"
     ]
    }
   ],
   "source": [
    "# Plot normalized confusion matrix\n",
    "plt.figure(figsize=(8,5))\n",
    "cmn_lr =plot_confusion_matrix(cnf_matrix_lr, classes=class_names, normalize=True,\n",
    "                      title='Normalized Logistic Regression Confusion matrix')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Gradient-boosted Tree Tuning\n",
    " - maxDepth (Maximum depth of the tree, default=5): [5, 10]\n",
    " - minInstancesPerNode (Minimum number of instances each child must have after split, default=1): [1,3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6154c983504042e39cfd453a17bf3a70",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# parameter grid\n",
    "paramGrid_gbt = ParamGridBuilder() \\\n",
    "    .addGrid(model_gbt.stages[-1].maxDepth, [5, 10]) \\\n",
    "    .addGrid(model_gbt.stages[-1].minInstancesPerNode,[1,3]) \\\n",
    "    .build()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2cd3ba8dc7d44c11a1dc6098df4c44b3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in thread cell_monitor-45:\n",
      "Traceback (most recent call last):\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 926, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/opt/conda/lib/python3.7/threading.py\", line 870, in run\n",
      "    self._target(*self._args, **self._kwargs)\n",
      "  File \"/opt/conda/lib/python3.7/site-packages/awseditorssparkmonitoringwidget-1.0-py3.7.egg/awseditorssparkmonitoringwidget/cellmonitor.py\", line 178, in cell_monitor\n",
      "    job_binned_stages[job_id][stage_id] = all_stages[stage_id]\n",
      "KeyError: 3696\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient-boosted Tree Model Tuning time (mins):  18.77\n",
      "Summary Stats\n",
      "Accuracy = 0.9700044769437397\n",
      "Precision = 0.9568151147098516\n",
      "Recall = 0.9118971061093247\n",
      "F1 Score = 0.9338162660520252"
     ]
    }
   ],
   "source": [
    "# tune the model\n",
    "cvmodel_gbt, cvprediction_gbt, cvtime_gbt = model_tuning(pipeline_gbt,paramGrid_gbt,train, test)\n",
    "print('Gradient-boosted Tree Model Tuning time (mins): ',cvtime_gbt)\n",
    "cvacc_gbt, cvf1_gbt = evaluation(cvprediction_gbt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e0c4a747603942caa0a20f2f77ef9904",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   maxDepth  minInstancesPerNode        f1\n",
      "3        10                    3  0.930644\n",
      "2        10                    1  0.925611\n",
      "0         5                    1  0.925429\n",
      "1         5                    3  0.925377"
     ]
    }
   ],
   "source": [
    "# training model results\n",
    "cvmodel_result(cvmodel_gbt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5d86750cce7e4a2fb18e841f3fe31bed",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#show feature importances\n",
    "feature_importance(cvmodel_gbt,cvmodel_gbt.bestModel.stages[-1].featureImportances) \\\n",
    ".plot(kind ='barh',x ='feature',y='importance',color ='navy',legend=None,\\\n",
    "      title = 'Gradient-boosted Tree Feature Importance (Full 12GB Dataset)',figsize = (8,5)).invert_yaxis();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2c8b5155de164186aa43be1b6b666bbc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "array([[5082,   64],\n",
      "       [ 137, 1418]])"
     ]
    }
   ],
   "source": [
    "#plot confusion_matrix\n",
    "y_truegbt = cvprediction_gbt.select(\"label\").withColumn(\"label\",col(\"label\").cast(DoubleType()))\n",
    "y_truegbt = y_truegbt.toPandas()\n",
    "y_predgbt = cvprediction_gbt.select(\"prediction\")\n",
    "y_predgbt = y_predgbt.toPandas()\n",
    "\n",
    "cnf_matrix = confusion_matrix(y_truegbt, y_predgbt)\n",
    "cnf_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9c53a0099a8e4d49a5cdc63847de18d7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[5082   64]\n",
      " [ 137 1418]]"
     ]
    }
   ],
   "source": [
    "# Plot non-normalized confusion matrix\n",
    "plt.figure(figsize=(8,5))\n",
    "cm_gbt = plot_confusion_matrix(cnf_matrix, classes=class_names,\n",
    "                      title='Gradient-boosted Tree Confusion matrix\\nwithout normalization')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7547824c5e504219857d0fe271951ff7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized confusion matrix\n",
      "[[0.98756316 0.01243684]\n",
      " [0.08810289 0.91189711]]"
     ]
    }
   ],
   "source": [
    "# Plot normalized confusion matrix\n",
    "plt.figure(figsize=(8,5))\n",
    "cmn_gbt =plot_confusion_matrix(cnf_matrix, classes=class_names, normalize=True,\n",
    "                      title='Normalized Gradient-boosted Tree Confusion matrix')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplot plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Random Forest Tuning\n",
    " - minInstancesPerNode (Minimum number of instances each child must have after split, default=1): [1,3]\n",
    " - numTrees (Number of trees in the random forest, default =20):[20,30]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# parameter grid\n",
    "paramGrid_rf = ParamGridBuilder() \\\n",
    "    .addGrid(model_rf.stages[-1].minInstancesPerNode,[1,3]) \\\n",
    "    .addGrid(model_rf.stages[-1].numTrees, [20, 30]) \\\n",
    "    .build()\n",
    "    \n",
    "# tune the model\n",
    "cvmodel_rf, cvprediction_rf, cvtime_rf = model_tuning(pipeline_rf,paramGrid_rf,train, test)\n",
    "print('Random Forest Model Tuning time (mins): ',cvtime_rf)\n",
    "cvacc_rf, cvf1_rf = evaluation(cvprediction_rf)    \n",
    "\n",
    "# model results\n",
    "print(cvmodel_result(cvmodel_rf))\n",
    "\n",
    "#show feature importances\n",
    "feature_importance(cvmodel_rf,cvmodel_rf.bestModel.stages[-1].featureImportances) \\\n",
    ".plot(kind ='bar',x ='feature',y='importance',color ='navy', \\\n",
    "      title = 'Random Forest Feature Importance');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Model Tuning Performance Summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c256fd418b724583ab624477bbc7997c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "VBox()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatProgress(value=0.0, bar_style='info', description='Progress:', layout=Layout(height='25px', width='50%'),…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Tuning Performance Summary\n",
      "\n",
      "Logistic Regression Base Model results: F1 Score 0.815, Accuracy is 0.92\n",
      "Logistic Regression Model Tuning results: F1 Score 0.827, Accuracy is 0.921\n",
      "Gradient-boosted Tree Base Model results: F1 Score 0.843, Accuracy is 0.933\n",
      "Gradient-boosted Tree Model Tuning results: F1 Score 0.934, Accuracy is 0.97"
     ]
    }
   ],
   "source": [
    "# Performance summary for Model Tuning\n",
    "print('Model Tuning Performance Summary\\n')\n",
    "print(\"Logistic Regression Base Model results: F1 Score {}, Accuracy is {}\".format(f1_lr, acc_lr))\n",
    "print(\"Logistic Regression Model Tuning results: F1 Score {}, Accuracy is {}\".format(cvf1_lr, cvacc_lr))\n",
    "print(\"Gradient-boosted Tree Base Model results: F1 Score {}, Accuracy is {}\".format(f1_gbt, acc_gbt))\n",
    "print(\"Gradient-boosted Tree Model Tuning results: F1 Score {}, Accuracy is {}\".format(cvf1_gbt, cvacc_gbt))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "| Models | F1 Score | Accuracy |\n",
    "| --- | --- | --- |\n",
    "| Initial Logistic Regression | 0.815 | 0.920|\n",
    "| Tuned Logistic Regression | 0.827 | 0.921|\n",
    "| Initial Gradient-boosted Tree | 0.843 | 0.933|\n",
    "| Tuned Gradient-boosted Tree | 0.934 | 0.970|"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "PySpark",
   "language": "",
   "name": "pysparkkernel"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "python",
    "version": 2
   },
   "mimetype": "text/x-python",
   "name": "pyspark",
   "pygments_lexer": "python2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}