RF_DT_GBT_LR_Binary_Bal.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ERROR:py4j.java_gateway:An error occurred while trying to connect to the Java server (127.0.0.1:62046)\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 958, in _get_connection\n",
      "    connection = self.deque.pop()\n",
      "IndexError: pop from an empty deque\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1096, in start\n",
      "    self.socket.connect((self.address, self.port))\n",
      "ConnectionRefusedError: [WinError 10061] No connection could be made because the target machine actively refused it\n"
     ]
    },
    {
     "ename": "Py4JNetworkError",
     "evalue": "An error occurred while trying to connect to the Java server (127.0.0.1:62046)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_get_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    957\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 958\u001b[1;33m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdeque\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    959\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mIndexError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mIndexError\u001b[0m: pop from an empty deque",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mConnectionRefusedError\u001b[0m                    Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36mstart\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m   1095\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1096\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msocket\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mconnect\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddress\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mport\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1097\u001b[0m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstream\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msocket\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmakefile\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"rb\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mConnectionRefusedError\u001b[0m: [WinError 10061] No connection could be made because the target machine actively refused it",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m                          Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-214-9a6bb26e06fe>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     24\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0mpyspark\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mml\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mevaluation\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mBinaryClassificationEvaluator\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     25\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0mpyspark\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mml\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfeature\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mOneHotEncoder\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mOneHotEncoderModel\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mStringIndexer\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mVectorAssembler\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 26\u001b[1;33m \u001b[0mspark\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mSparkSession\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbuilder\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mgetOrCreate\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     27\u001b[0m \u001b[0msc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mspark\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msparkContext\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\sql\\session.py\u001b[0m in \u001b[0;36mgetOrCreate\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    186\u001b[0m                     \u001b[0msession\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mSparkSession\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msc\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    187\u001b[0m                 \u001b[1;32mfor\u001b[0m \u001b[0mkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalue\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_options\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 188\u001b[1;33m                     \u001b[0msession\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_jsparkSession\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msessionState\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mconf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msetConfString\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    189\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0msession\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    190\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m__call__\u001b[1;34m(self, *args)\u001b[0m\n\u001b[0;32m   1282\u001b[0m             \u001b[0mproto\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mEND_COMMAND_PART\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1283\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1284\u001b[1;33m         \u001b[0manswer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mgateway_client\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1285\u001b[0m         return_value = get_return_value(\n\u001b[0;32m   1286\u001b[0m             answer, self.gateway_client, self.target_id, self.name)\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command, retry, binary)\u001b[0m\n\u001b[0;32m   1010\u001b[0m          \u001b[1;32mif\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0mbinary\u001b[0m\u001b[0;31m`\u001b[0m \u001b[1;32mis\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;33m.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1011\u001b[0m         \"\"\"\n\u001b[1;32m-> 1012\u001b[1;33m         \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1013\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1014\u001b[0m             \u001b[0mresponse\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_get_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    958\u001b[0m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdeque\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    959\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mIndexError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 960\u001b[1;33m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_create_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    961\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    962\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_create_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    964\u001b[0m         connection = GatewayConnection(\n\u001b[0;32m    965\u001b[0m             self.gateway_parameters, self.gateway_property)\n\u001b[1;32m--> 966\u001b[1;33m         \u001b[0mconnection\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstart\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    967\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    968\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36mstart\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m   1106\u001b[0m                 \u001b[1;34m\"server ({0}:{1})\"\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddress\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mport\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1107\u001b[0m             \u001b[0mlogger\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexception\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1108\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mPy4JNetworkError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1109\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1110\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_authenticate_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m: An error occurred while trying to connect to the Java server (127.0.0.1:62046)"
     ]
    }
   ],
   "source": [
    "from pyspark.sql import SparkSession\n",
    "from pyspark.sql import Row\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from pyspark.sql.types import *\n",
    "from pyspark.sql.functions import *\n",
    "import matplotlib.pyplot as plt\n",
    "from pyspark.sql import functions as fn\n",
    "from pyspark.ml import feature, regression, evaluation, Pipeline\n",
    "import seaborn as sns\n",
    "from pyspark.ml.feature import VectorAssembler\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.regression import LinearRegression\n",
    "from pyspark.ml.stat import Correlation\n",
    "from pyspark.ml.feature import VectorAssembler\n",
    "from pyspark.ml.classification import RandomForestClassifier\n",
    "from pyspark.ml.classification import LogisticRegression,RandomForestClassifier\n",
    "from pyspark.ml.classification import GBTClassifier\n",
    "from pyspark.ml.evaluation import BinaryClassificationEvaluator\n",
    "from pyspark.ml.classification import DecisionTreeClassifier\n",
    "from pyspark.ml.tuning import CrossValidator, ParamGridBuilder\n",
    "from pyspark.ml.classification import LogisticRegression\n",
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.evaluation import BinaryClassificationEvaluator\n",
    "from pyspark.ml.feature import OneHotEncoder, OneHotEncoderModel, StringIndexer, VectorAssembler\n",
    "spark = SparkSession.builder.getOrCreate()\n",
    "sc = spark.sparkContext\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Do not delete or change this cell\n",
    "\n",
    "import os\n",
    "\n",
    "# Define a function to determine if we are running on data bricks\n",
    "# Return true if running in the data bricks environment, false otherwise\n",
    "def is_databricks():\n",
    "    # get the databricks runtime version\n",
    "    db_env = os.getenv(\"DATABRICKS_RUNTIME_VERSION\")\n",
    "    \n",
    "    # if running on data bricks\n",
    "    if db_env != None:\n",
    "        return True\n",
    "    else:\n",
    "        return False\n",
    "\n",
    "# Define a function to read the data file.  The full path data file name is constructed\n",
    "# by checking runtime environment variables to determine if the runtime environment is \n",
    "# databricks, or a student's personal computer.  The full path file name is then\n",
    "# constructed based on the runtime env.\n",
    "# \n",
    "# Params\n",
    "#   data_file_name: The base name of the data file to load\n",
    "# \n",
    "# Returns the full path file name based on the runtime env\n",
    "#\n",
    "def get_training_filename(data_file_name):    \n",
    "    # if running on data bricks\n",
    "    if is_databricks():\n",
    "        # build the full path file name assuming data brick env\n",
    "        full_path_name = \"/FileStore/tables/%s\" % data_file_name\n",
    "    # else the data is assumed to be in the same dir as this notebook\n",
    "    else:\n",
    "        # Assume the student is running on their own computer and load the data\n",
    "        # file from the same dir as this notebook\n",
    "        full_path_name = data_file_name\n",
    "    \n",
    "    # return the full path file name to the caller\n",
    "    return full_path_name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Loading Train data\n",
    "\n",
    "us_train_cat = spark.read.csv(get_training_filename('USAccident_train_bal_bin.csv'), header = True, inferSchema = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------------+\n",
      "|count(Severity)|\n",
      "+---------------+\n",
      "|              2|\n",
      "+---------------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Number of unique labels in Severity column\n",
    "\n",
    "us_train_cat.agg(countDistinct(\"Severity\")).show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+------+\n",
      "|Severity| count|\n",
      "+--------+------+\n",
      "|       1|258836|\n",
      "|       0|263700|\n",
      "+--------+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Checking the balance of data in training dataset\n",
    "\n",
    "us_train_cat.groupBy('Severity').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Loading the test data\n",
    "\n",
    "us_test_cat = spark.read.csv(get_training_filename('USAccident_validation_new.csv'), header = True, inferSchema = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------+------+\n",
      "|Severity| count|\n",
      "+--------+------+\n",
      "|       3| 58617|\n",
      "|       4|  5993|\n",
      "|       2|131790|\n",
      "+--------+------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Checking the balance of data in testing dataset\n",
    "\n",
    "us_test_cat.groupBy('Severity').count().show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Vector Assembler to convert all features except Severity to a single column features for feeding it to input of model\n",
    "\n",
    "va = VectorAssembler().setInputCols([i for i in us_train_cat.columns if i!='Severity']).setOutputCol('features')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# String Indexer to assign target Variable Severity name Label needed for the model to predict\n",
    "\n",
    "label_stringIdx = StringIndexer(inputCol=\"Severity\", outputCol=\"label\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Assigning label of 1 to category 3 and 4 and combine them to make it 1 category for train data\n",
    "\n",
    "us_train_cat=us_train_cat.withColumn(\"Severity\",when(((us_train_cat[\"Severity\"]==4) | (us_train_cat[\"Severity\"]==3)),1).otherwise(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Assigning label of 1 to category 3 and 4 and combine them to make it 1 category for test data\n",
    "\n",
    "us_test_cat=us_test_cat.withColumn(\"Severity\",when(((us_test_cat[\"Severity\"]==4) | (us_test_cat[\"Severity\"]==3)),1).otherwise(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluator defined for Binary Classification\n",
    "\n",
    "evaluator_rfb = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RF Base Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Create an initial RandomForest model.\n",
    "rf = RandomForestClassifier(labelCol=\"label\", featuresCol=\"features\",seed=42)\n",
    "\n",
    "# Pipeline with stages for fitting the training data\n",
    "\n",
    "rfModel = Pipeline(stages=[label_stringIdx,va, rf])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fit the training data using RF pipeline\n",
    "\n",
    "rf_fit = rfModel.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Predict the test data using fitted train pipeline\n",
    "\n",
    "pred_rfbb = rf_fit.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.7516960347127739\n"
     ]
    }
   ],
   "source": [
    "# AUC Score for the test data\n",
    "\n",
    "print(\"AUC Score is\", evaluator_rfb.evaluate(pred_rfbb))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_rfbb=(pred_rfbb).toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.59      0.70    131790\n",
      "           1       0.49      0.80      0.61     64610\n",
      "\n",
      "    accuracy                           0.66    196400\n",
      "   macro avg       0.68      0.70      0.66    196400\n",
      "weighted avg       0.74      0.66      0.67    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_rfbb,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SparseVector(119, {2: 0.0001, 3: 0.0001, 4: 0.0, 5: 0.0, 7: 0.0, 9: 0.0006, 13: 0.0003, 14: 0.0001, 15: 0.0, 17: 0.0, 19: 0.0, 20: 0.0, 21: 0.0, 23: 0.0228, 24: 0.0088, 25: 0.0142, 26: 0.0025, 27: 0.006, 28: 0.0, 30: 0.0268, 32: 0.0003, 34: 0.0164, 36: 0.0976, 38: 0.0, 40: 0.0007, 42: 0.0017, 44: 0.0025, 48: 0.0, 49: 0.0, 50: 0.0014, 51: 0.0508, 52: 0.0274, 53: 0.0, 54: 0.0, 55: 0.0, 58: 0.0001, 59: 0.0, 60: 0.0001, 61: 0.0, 62: 0.0001, 63: 0.0, 65: 0.0, 66: 0.0, 67: 0.0, 68: 0.0, 70: 0.0, 71: 0.0001, 73: 0.0, 75: 0.0001, 77: 0.0, 78: 0.0, 81: 0.1403, 82: 0.0434, 83: 0.0004, 84: 0.0001, 85: 0.0, 87: 0.0, 89: 0.0, 92: 0.0, 93: 0.0, 94: 0.0, 95: 0.0112, 97: 0.0, 98: 0.0005, 100: 0.05, 101: 0.0002, 102: 0.0003, 103: 0.0014, 104: 0.0, 105: 0.0209, 106: 0.0001, 107: 0.0121, 109: 0.1052, 110: 0.0, 111: 0.0209, 113: 0.001, 115: 0.0088, 116: 0.013, 117: 0.0, 118: 0.288})"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Feature Importance from RF model \n",
    "\n",
    "rf_fit.stages[-1].featureImportances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of RF Base Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_rfbb = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], rf_fit.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from Random Forest binary balanced')"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of RF Base Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_rfbb['column'][:10], y=feat_imp_tuned_rfbb['weight'][:10],data=feat_imp_tuned_rfbb)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from Random Forest binary balanced\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6618991853360489"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Accuracy calculation for RF Base Model\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]\n",
    "\n",
    "binary_prediction=(pred_rfbb).select(\"prediction\").collect()\n",
    "\n",
    "binary_true_labels=us_test_cat.select(\"Severity\").collect()\n",
    "\n",
    "np.sum(list([int(binary_true_labels[i][0]==binary_prediction[i][0]) for i in range(len(true_labels))]))/len(true_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RF Grid Search "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ERROR:py4j.java_gateway:An error occurred while trying to connect to the Java server (127.0.0.1:62046)\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1174, in send_command\n",
      "    self.socket.sendall(command.encode(\"utf-8\"))\n",
      "ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1014, in send_command\n",
      "    response = connection.send_command(command)\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1178, in send_command\n",
      "    \"Error while sending\", e, proto.ERROR_ON_SEND)\n",
      "py4j.protocol.Py4JNetworkError: Error while sending\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 958, in _get_connection\n",
      "    connection = self.deque.pop()\n",
      "IndexError: pop from an empty deque\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1096, in start\n",
      "    self.socket.connect((self.address, self.port))\n",
      "ConnectionRefusedError: [WinError 10061] No connection could be made because the target machine actively refused it\n"
     ]
    },
    {
     "ename": "Py4JNetworkError",
     "evalue": "An error occurred while trying to connect to the Java server (127.0.0.1:62046)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mConnectionResetError\u001b[0m                      Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command)\u001b[0m\n\u001b[0;32m   1173\u001b[0m             \u001b[1;31m# if it sent a RST packet (SO_LINGER)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1174\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msocket\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msendall\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mencode\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"utf-8\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1175\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mConnectionResetError\u001b[0m: [WinError 10054] An existing connection was forcibly closed by the remote host",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m                          Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command, retry, binary)\u001b[0m\n\u001b[0;32m   1013\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1014\u001b[1;33m             \u001b[0mresponse\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1015\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mbinary\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command)\u001b[0m\n\u001b[0;32m   1177\u001b[0m             raise Py4JNetworkError(\n\u001b[1;32m-> 1178\u001b[1;33m                 \"Error while sending\", e, proto.ERROR_ON_SEND)\n\u001b[0m\u001b[0;32m   1179\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m: Error while sending",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_get_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    957\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 958\u001b[1;33m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdeque\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    959\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mIndexError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mIndexError\u001b[0m: pop from an empty deque",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mConnectionRefusedError\u001b[0m                    Traceback (most recent call last)",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36mstart\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m   1095\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1096\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msocket\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mconnect\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddress\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mport\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1097\u001b[0m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstream\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msocket\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmakefile\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"rb\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mConnectionRefusedError\u001b[0m: [WinError 10061] No connection could be made because the target machine actively refused it",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m                          Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-210-2160e113b8e0>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m# Create an initial RandomForest model.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0mrf_new\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mRandomForestClassifier\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlabelCol\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m\"label\"\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfeaturesCol\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m\"features\"\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mseed\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m42\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;31m# Train model with Training Data\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mrfModel_new\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mPipeline\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mstages\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mlabel_stringIdx\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mva\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mrf_new\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\__init__.py\u001b[0m in \u001b[0;36mwrapper\u001b[1;34m(self, *args, **kwargs)\u001b[0m\n\u001b[0;32m    109\u001b[0m             \u001b[1;32mraise\u001b[0m \u001b[0mTypeError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Method %s forces keyword arguments.\"\u001b[0m \u001b[1;33m%\u001b[0m \u001b[0mfunc\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__name__\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    110\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_input_kwargs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 111\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    112\u001b[0m     \u001b[1;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    113\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\classification.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, featuresCol, labelCol, predictionCol, probabilityCol, rawPredictionCol, maxDepth, maxBins, minInstancesPerNode, minInfoGain, maxMemoryInMB, cacheNodeIds, checkpointInterval, impurity, numTrees, featureSubsetStrategy, seed, subsamplingRate, leafCol, minWeightFractionPerNode)\u001b[0m\n\u001b[0;32m   1424\u001b[0m         \u001b[0msuper\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mRandomForestClassifier\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__init__\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1425\u001b[0m         self._java_obj = self._new_java_obj(\n\u001b[1;32m-> 1426\u001b[1;33m             \"org.apache.spark.ml.classification.RandomForestClassifier\", self.uid)\n\u001b[0m\u001b[0;32m   1427\u001b[0m         self._setDefault(maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,\n\u001b[0;32m   1428\u001b[0m                          \u001b[0mmaxMemoryInMB\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m256\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcacheNodeIds\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcheckpointInterval\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m10\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\wrapper.py\u001b[0m in \u001b[0;36m_new_java_obj\u001b[1;34m(java_class, *args)\u001b[0m\n\u001b[0;32m     65\u001b[0m         \u001b[0mjava_obj\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_jvm\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     66\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mname\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mjava_class\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msplit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\".\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 67\u001b[1;33m             \u001b[0mjava_obj\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mjava_obj\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mname\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     68\u001b[0m         \u001b[0mjava_args\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[0m_py2java\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msc\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0marg\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0marg\u001b[0m \u001b[1;32min\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     69\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mjava_obj\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mjava_args\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m__getattr__\u001b[1;34m(self, name)\u001b[0m\n\u001b[0;32m   1676\u001b[0m             \u001b[0mproto\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mREFLECTION_COMMAND_NAME\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1677\u001b[0m             \u001b[0mproto\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mREFL_GET_UNKNOWN_SUB_COMMAND_NAME\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mname\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;34m\"\\n\"\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_id\u001b[0m \u001b[1;33m+\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1678\u001b[1;33m             \"\\n\" + proto.END_COMMAND_PART)\n\u001b[0m\u001b[0;32m   1679\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0manswer\u001b[0m \u001b[1;33m==\u001b[0m \u001b[0mproto\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mSUCCESS_PACKAGE\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1680\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mJavaPackage\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mname\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_gateway_client\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mjvm_id\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_id\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command, retry, binary)\u001b[0m\n\u001b[0;32m   1027\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_should_retry\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mretry\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mpne\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1028\u001b[0m                 \u001b[0mlogging\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0minfo\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Exception while sending command.\"\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mexc_info\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1029\u001b[1;33m                 \u001b[0mresponse\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbinary\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mbinary\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1030\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1031\u001b[0m                 logging.exception(\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36msend_command\u001b[1;34m(self, command, retry, binary)\u001b[0m\n\u001b[0;32m   1010\u001b[0m          \u001b[1;32mif\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0mbinary\u001b[0m\u001b[0;31m`\u001b[0m \u001b[1;32mis\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;33m.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1011\u001b[0m         \"\"\"\n\u001b[1;32m-> 1012\u001b[1;33m         \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1013\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1014\u001b[0m             \u001b[0mresponse\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_get_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    958\u001b[0m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdeque\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpop\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    959\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mIndexError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 960\u001b[1;33m             \u001b[0mconnection\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_create_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    961\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    962\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m_create_connection\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    964\u001b[0m         connection = GatewayConnection(\n\u001b[0;32m    965\u001b[0m             self.gateway_parameters, self.gateway_property)\n\u001b[1;32m--> 966\u001b[1;33m         \u001b[0mconnection\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstart\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    967\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mconnection\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    968\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36mstart\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m   1106\u001b[0m                 \u001b[1;34m\"server ({0}:{1})\"\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddress\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mport\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1107\u001b[0m             \u001b[0mlogger\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexception\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1108\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mPy4JNetworkError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1109\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1110\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_authenticate_connection\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mPy4JNetworkError\u001b[0m: An error occurred while trying to connect to the Java server (127.0.0.1:62046)"
     ]
    }
   ],
   "source": [
    "\n",
    "# Create an initial RandomForest model\n",
    "\n",
    "rf_new = RandomForestClassifier(labelCol=\"label\", featuresCol=\"features\",seed=42)\n",
    "\n",
    "# Pipeline with stages for RF Grid Search Model\n",
    "\n",
    "rfModel_new = Pipeline(stages=[label_stringIdx,va, rf_new])\n",
    "\n",
    "# Grid Search for tuning Hyper parameters \n",
    "\n",
    "paramGrid_rft = ParamGridBuilder().addGrid(rf_new.numTrees, [10,25,60]).addGrid(rf_new.maxDepth, [3,6,10]).addGrid(rf_new.impurity,[\"entropy\", \"gini\"]).build()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 209,
   "metadata": {},
   "outputs": [
    {
     "ename": "Py4JJavaError",
     "evalue": "An error occurred while calling o124422.fit.\n: org.apache.spark.SparkException: Job aborted due to stage failure: Task 7 in stage 51187.0 failed 1 times, most recent failure: Lost task 7.0 in stage 51187.0 (TID 368623, DESKTOP-TT8TT9T.fios-router.home, executor driver): java.lang.OutOfMemoryError: Java heap space\r\n\tat java.lang.reflect.Array.newArray(Native Method)\r\n\tat java.lang.reflect.Array.newInstance(Unknown Source)\r\n\tat java.io.ObjectInputStream.readArray(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:76)\r\n\tat org.apache.spark.serializer.DeserializationStream$$anon$1.getNext(Serializer.scala:168)\r\n\tat org.apache.spark.util.NextIterator.hasNext(NextIterator.scala:73)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIterator(MemoryStore.scala:221)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIteratorAsValues(MemoryStore.scala:299)\r\n\tat org.apache.spark.storage.BlockManager.maybeCacheDiskValuesInMemory(BlockManager.scala:1516)\r\n\tat org.apache.spark.storage.BlockManager.getLocalValues(BlockManager.scala:825)\r\n\tat org.apache.spark.storage.BlockManager.get(BlockManager.scala:1111)\r\n\tat org.apache.spark.storage.BlockManager.getOrElseUpdate(BlockManager.scala:1178)\r\n\tat org.apache.spark.rdd.RDD.getOrCompute(RDD.scala:360)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:311)\r\n\tat org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)\r\n\tat org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:313)\r\n\tat org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:52)\r\n\tat org.apache.spark.scheduler.Task.run(Task.scala:127)\r\n\nDriver stacktrace:\r\n\tat org.apache.spark.scheduler.DAGScheduler.failJobAndIndependentStages(DAGScheduler.scala:1989)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$abortStage$2(DAGScheduler.scala:1977)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$abortStage$2$adapted(DAGScheduler.scala:1976)\r\n\tat scala.collection.mutable.ResizableArray.foreach(ResizableArray.scala:62)\r\n\tat scala.collection.mutable.ResizableArray.foreach$(ResizableArray.scala:55)\r\n\tat scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:49)\r\n\tat org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1976)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$handleTaskSetFailed$1(DAGScheduler.scala:956)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$handleTaskSetFailed$1$adapted(DAGScheduler.scala:956)\r\n\tat scala.Option.foreach(Option.scala:407)\r\n\tat org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:956)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:2206)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:2155)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:2144)\r\n\tat org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:49)\r\n\tat org.apache.spark.scheduler.DAGScheduler.runJob(DAGScheduler.scala:758)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2116)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2137)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2156)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2181)\r\n\tat org.apache.spark.rdd.RDD.$anonfun$collect$1(RDD.scala:1004)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)\r\n\tat org.apache.spark.rdd.RDD.withScope(RDD.scala:388)\r\n\tat org.apache.spark.rdd.RDD.collect(RDD.scala:1003)\r\n\tat org.apache.spark.rdd.PairRDDFunctions.$anonfun$collectAsMap$1(PairRDDFunctions.scala:737)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)\r\n\tat org.apache.spark.rdd.RDD.withScope(RDD.scala:388)\r\n\tat org.apache.spark.rdd.PairRDDFunctions.collectAsMap(PairRDDFunctions.scala:736)\r\n\tat org.apache.spark.ml.tree.impl.RandomForest$.findBestSplits(RandomForest.scala:588)\r\n\tat org.apache.spark.ml.tree.impl.RandomForest$.run(RandomForest.scala:226)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.$anonfun$train$1(RandomForestClassifier.scala:144)\r\n\tat org.apache.spark.ml.util.Instrumentation$.$anonfun$instrumented$1(Instrumentation.scala:191)\r\n\tat scala.util.Try$.apply(Try.scala:213)\r\n\tat org.apache.spark.ml.util.Instrumentation$.instrumented(Instrumentation.scala:191)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.train(RandomForestClassifier.scala:122)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.train(RandomForestClassifier.scala:48)\r\n\tat org.apache.spark.ml.Predictor.fit(Predictor.scala:152)\r\n\tat sun.reflect.GeneratedMethodAccessor2715.invoke(Unknown Source)\r\n\tat sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)\r\n\tat java.lang.reflect.Method.invoke(Unknown Source)\r\n\tat py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)\r\n\tat py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)\r\n\tat py4j.Gateway.invoke(Gateway.java:282)\r\n\tat py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)\r\n\tat py4j.commands.CallCommand.execute(CallCommand.java:79)\r\n\tat py4j.GatewayConnection.run(GatewayConnection.java:238)\r\n\tat java.lang.Thread.run(Unknown Source)\r\nCaused by: java.lang.OutOfMemoryError: Java heap space\r\n\tat java.lang.reflect.Array.newArray(Native Method)\r\n\tat java.lang.reflect.Array.newInstance(Unknown Source)\r\n\tat java.io.ObjectInputStream.readArray(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:76)\r\n\tat org.apache.spark.serializer.DeserializationStream$$anon$1.getNext(Serializer.scala:168)\r\n\tat org.apache.spark.util.NextIterator.hasNext(NextIterator.scala:73)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIterator(MemoryStore.scala:221)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIteratorAsValues(MemoryStore.scala:299)\r\n\tat org.apache.spark.storage.BlockManager.maybeCacheDiskValuesInMemory(BlockManager.scala:1516)\r\n\tat org.apache.spark.storage.BlockManager.getLocalValues(BlockManager.scala:825)\r\n\tat org.apache.spark.storage.BlockManager.get(BlockManager.scala:1111)\r\n\tat org.apache.spark.storage.BlockManager.getOrElseUpdate(BlockManager.scala:1178)\r\n\tat org.apache.spark.rdd.RDD.getOrCompute(RDD.scala:360)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:311)\r\n\tat org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)\r\n\tat org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:313)\r\n\tat org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:52)\r\n\tat org.apache.spark.scheduler.Task.run(Task.scala:127)\r\n",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mPy4JJavaError\u001b[0m                             Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-209-922a149398e7>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mcv_rf\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mCrossValidator\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mestimator\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mrfModel_new\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mestimatorParamMaps\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mparamGrid_rft\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mevaluator\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mevaluator_rfb\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mnumFolds\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m5\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mseed\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m42\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mus_train_cat\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\base.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, dataset, params)\u001b[0m\n\u001b[0;32m    129\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcopy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparams\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    130\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 131\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    132\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    133\u001b[0m             raise ValueError(\"Params must be either a param map or a list/tuple of param maps, \"\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\tuning.py\u001b[0m in \u001b[0;36m_fit\u001b[1;34m(self, dataset)\u001b[0m\n\u001b[0;32m    350\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    351\u001b[0m             \u001b[0mtasks\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_parallelFitTasks\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mest\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0meva\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalidation\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mepm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcollectSubModelsParam\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 352\u001b[1;33m             \u001b[1;32mfor\u001b[0m \u001b[0mj\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmetric\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msubModel\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mpool\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mimap_unordered\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;32mlambda\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtasks\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    353\u001b[0m                 \u001b[0mmetrics\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mj\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mmetric\u001b[0m \u001b[1;33m/\u001b[0m \u001b[0mnFolds\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    354\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[0mcollectSubModelsParam\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\multiprocessing\\pool.py\u001b[0m in \u001b[0;36mnext\u001b[1;34m(self, timeout)\u001b[0m\n\u001b[0;32m    746\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0msuccess\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    747\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mvalue\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 748\u001b[1;33m         \u001b[1;32mraise\u001b[0m \u001b[0mvalue\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    749\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    750\u001b[0m     \u001b[0m__next__\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnext\u001b[0m                    \u001b[1;31m# XXX\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\multiprocessing\\pool.py\u001b[0m in \u001b[0;36mworker\u001b[1;34m(inqueue, outqueue, initializer, initargs, maxtasks, wrap_exception)\u001b[0m\n\u001b[0;32m    119\u001b[0m         \u001b[0mjob\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mi\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfunc\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkwds\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtask\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    120\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 121\u001b[1;33m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m(\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfunc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    122\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    123\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mwrap_exception\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mfunc\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0m_helper_reraises_exception\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\tuning.py\u001b[0m in \u001b[0;36m<lambda>\u001b[1;34m(f)\u001b[0m\n\u001b[0;32m    350\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    351\u001b[0m             \u001b[0mtasks\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_parallelFitTasks\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mest\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0meva\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalidation\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mepm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcollectSubModelsParam\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 352\u001b[1;33m             \u001b[1;32mfor\u001b[0m \u001b[0mj\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmetric\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msubModel\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mpool\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mimap_unordered\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;32mlambda\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtasks\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    353\u001b[0m                 \u001b[0mmetrics\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mj\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mmetric\u001b[0m \u001b[1;33m/\u001b[0m \u001b[0mnFolds\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    354\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[0mcollectSubModelsParam\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\tuning.py\u001b[0m in \u001b[0;36msingleTask\u001b[1;34m()\u001b[0m\n\u001b[0;32m     50\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     51\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0msingleTask\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 52\u001b[1;33m         \u001b[0mindex\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodelIter\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     53\u001b[0m         \u001b[0mmetric\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0meva\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtransform\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mvalidation\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mepm\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     54\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mindex\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmetric\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;32mif\u001b[0m \u001b[0mcollectSubModel\u001b[0m \u001b[1;32melse\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\base.py\u001b[0m in \u001b[0;36m__next__\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m     60\u001b[0m                 \u001b[1;32mraise\u001b[0m \u001b[0mStopIteration\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"No models remaining.\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     61\u001b[0m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcounter\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 62\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mindex\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfitSingleModel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     63\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     64\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mnext\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\base.py\u001b[0m in \u001b[0;36mfitSingleModel\u001b[1;34m(index)\u001b[0m\n\u001b[0;32m    103\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    104\u001b[0m         \u001b[1;32mdef\u001b[0m \u001b[0mfitSingleModel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 105\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mestimator\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mparamMaps\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    106\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    107\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0m_FitMultipleIterator\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfitSingleModel\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparamMaps\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\base.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, dataset, params)\u001b[0m\n\u001b[0;32m    127\u001b[0m         \u001b[1;32melif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparams\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdict\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    128\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mparams\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 129\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcopy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparams\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    130\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    131\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\pipeline.py\u001b[0m in \u001b[0;36m_fit\u001b[1;34m(self, dataset)\u001b[0m\n\u001b[0;32m    107\u001b[0m                     \u001b[0mdataset\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mstage\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtransform\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    108\u001b[0m                 \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m  \u001b[1;31m# must be an Estimator\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 109\u001b[1;33m                     \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mstage\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    110\u001b[0m                     \u001b[0mtransformers\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    111\u001b[0m                     \u001b[1;32mif\u001b[0m \u001b[0mi\u001b[0m \u001b[1;33m<\u001b[0m \u001b[0mindexOfLastEstimator\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\base.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, dataset, params)\u001b[0m\n\u001b[0;32m    129\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcopy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparams\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    130\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 131\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    132\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    133\u001b[0m             raise ValueError(\"Params must be either a param map or a list/tuple of param maps, \"\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\wrapper.py\u001b[0m in \u001b[0;36m_fit\u001b[1;34m(self, dataset)\u001b[0m\n\u001b[0;32m    319\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    320\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 321\u001b[1;33m         \u001b[0mjava_model\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fit_java\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    322\u001b[0m         \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_create_model\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mjava_model\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    323\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_copyValues\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\ml\\wrapper.py\u001b[0m in \u001b[0;36m_fit_java\u001b[1;34m(self, dataset)\u001b[0m\n\u001b[0;32m    316\u001b[0m         \"\"\"\n\u001b[0;32m    317\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_transfer_params_to_java\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 318\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_java_obj\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_jdf\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    319\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    320\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_fit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdataset\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\u001b[0m in \u001b[0;36m__call__\u001b[1;34m(self, *args)\u001b[0m\n\u001b[0;32m   1284\u001b[0m         \u001b[0manswer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mgateway_client\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msend_command\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mcommand\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1285\u001b[0m         return_value = get_return_value(\n\u001b[1;32m-> 1286\u001b[1;33m             answer, self.gateway_client, self.target_id, self.name)\n\u001b[0m\u001b[0;32m   1287\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1288\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mtemp_arg\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mtemp_args\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\pyspark\\sql\\utils.py\u001b[0m in \u001b[0;36mdeco\u001b[1;34m(*a, **kw)\u001b[0m\n\u001b[0;32m     96\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mdeco\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0ma\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkw\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     97\u001b[0m         \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 98\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mf\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0ma\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkw\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     99\u001b[0m         \u001b[1;32mexcept\u001b[0m \u001b[0mpy4j\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mprotocol\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mPy4JJavaError\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    100\u001b[0m             \u001b[0mconverted\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mconvert_exception\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0me\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mjava_exception\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\protocol.py\u001b[0m in \u001b[0;36mget_return_value\u001b[1;34m(answer, gateway_client, target_id, name)\u001b[0m\n\u001b[0;32m    326\u001b[0m                 raise Py4JJavaError(\n\u001b[0;32m    327\u001b[0m                     \u001b[1;34m\"An error occurred while calling {0}{1}{2}.\\n\"\u001b[0m\u001b[1;33m.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 328\u001b[1;33m                     format(target_id, \".\", name), value)\n\u001b[0m\u001b[0;32m    329\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    330\u001b[0m                 raise Py4JError(\n",
      "\u001b[1;31mPy4JJavaError\u001b[0m: An error occurred while calling o124422.fit.\n: org.apache.spark.SparkException: Job aborted due to stage failure: Task 7 in stage 51187.0 failed 1 times, most recent failure: Lost task 7.0 in stage 51187.0 (TID 368623, DESKTOP-TT8TT9T.fios-router.home, executor driver): java.lang.OutOfMemoryError: Java heap space\r\n\tat java.lang.reflect.Array.newArray(Native Method)\r\n\tat java.lang.reflect.Array.newInstance(Unknown Source)\r\n\tat java.io.ObjectInputStream.readArray(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:76)\r\n\tat org.apache.spark.serializer.DeserializationStream$$anon$1.getNext(Serializer.scala:168)\r\n\tat org.apache.spark.util.NextIterator.hasNext(NextIterator.scala:73)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIterator(MemoryStore.scala:221)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIteratorAsValues(MemoryStore.scala:299)\r\n\tat org.apache.spark.storage.BlockManager.maybeCacheDiskValuesInMemory(BlockManager.scala:1516)\r\n\tat org.apache.spark.storage.BlockManager.getLocalValues(BlockManager.scala:825)\r\n\tat org.apache.spark.storage.BlockManager.get(BlockManager.scala:1111)\r\n\tat org.apache.spark.storage.BlockManager.getOrElseUpdate(BlockManager.scala:1178)\r\n\tat org.apache.spark.rdd.RDD.getOrCompute(RDD.scala:360)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:311)\r\n\tat org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)\r\n\tat org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:313)\r\n\tat org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:52)\r\n\tat org.apache.spark.scheduler.Task.run(Task.scala:127)\r\n\nDriver stacktrace:\r\n\tat org.apache.spark.scheduler.DAGScheduler.failJobAndIndependentStages(DAGScheduler.scala:1989)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$abortStage$2(DAGScheduler.scala:1977)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$abortStage$2$adapted(DAGScheduler.scala:1976)\r\n\tat scala.collection.mutable.ResizableArray.foreach(ResizableArray.scala:62)\r\n\tat scala.collection.mutable.ResizableArray.foreach$(ResizableArray.scala:55)\r\n\tat scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:49)\r\n\tat org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1976)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$handleTaskSetFailed$1(DAGScheduler.scala:956)\r\n\tat org.apache.spark.scheduler.DAGScheduler.$anonfun$handleTaskSetFailed$1$adapted(DAGScheduler.scala:956)\r\n\tat scala.Option.foreach(Option.scala:407)\r\n\tat org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:956)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:2206)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:2155)\r\n\tat org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:2144)\r\n\tat org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:49)\r\n\tat org.apache.spark.scheduler.DAGScheduler.runJob(DAGScheduler.scala:758)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2116)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2137)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2156)\r\n\tat org.apache.spark.SparkContext.runJob(SparkContext.scala:2181)\r\n\tat org.apache.spark.rdd.RDD.$anonfun$collect$1(RDD.scala:1004)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)\r\n\tat org.apache.spark.rdd.RDD.withScope(RDD.scala:388)\r\n\tat org.apache.spark.rdd.RDD.collect(RDD.scala:1003)\r\n\tat org.apache.spark.rdd.PairRDDFunctions.$anonfun$collectAsMap$1(PairRDDFunctions.scala:737)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)\r\n\tat org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)\r\n\tat org.apache.spark.rdd.RDD.withScope(RDD.scala:388)\r\n\tat org.apache.spark.rdd.PairRDDFunctions.collectAsMap(PairRDDFunctions.scala:736)\r\n\tat org.apache.spark.ml.tree.impl.RandomForest$.findBestSplits(RandomForest.scala:588)\r\n\tat org.apache.spark.ml.tree.impl.RandomForest$.run(RandomForest.scala:226)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.$anonfun$train$1(RandomForestClassifier.scala:144)\r\n\tat org.apache.spark.ml.util.Instrumentation$.$anonfun$instrumented$1(Instrumentation.scala:191)\r\n\tat scala.util.Try$.apply(Try.scala:213)\r\n\tat org.apache.spark.ml.util.Instrumentation$.instrumented(Instrumentation.scala:191)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.train(RandomForestClassifier.scala:122)\r\n\tat org.apache.spark.ml.classification.RandomForestClassifier.train(RandomForestClassifier.scala:48)\r\n\tat org.apache.spark.ml.Predictor.fit(Predictor.scala:152)\r\n\tat sun.reflect.GeneratedMethodAccessor2715.invoke(Unknown Source)\r\n\tat sun.reflect.DelegatingMethodAccessorImpl.invoke(Unknown Source)\r\n\tat java.lang.reflect.Method.invoke(Unknown Source)\r\n\tat py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)\r\n\tat py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)\r\n\tat py4j.Gateway.invoke(Gateway.java:282)\r\n\tat py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)\r\n\tat py4j.commands.CallCommand.execute(CallCommand.java:79)\r\n\tat py4j.GatewayConnection.run(GatewayConnection.java:238)\r\n\tat java.lang.Thread.run(Unknown Source)\r\nCaused by: java.lang.OutOfMemoryError: Java heap space\r\n\tat java.lang.reflect.Array.newArray(Native Method)\r\n\tat java.lang.reflect.Array.newInstance(Unknown Source)\r\n\tat java.io.ObjectInputStream.readArray(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.defaultReadFields(Unknown Source)\r\n\tat java.io.ObjectInputStream.readSerialData(Unknown Source)\r\n\tat java.io.ObjectInputStream.readOrdinaryObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject0(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat java.io.ObjectInputStream.readObject(Unknown Source)\r\n\tat org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:76)\r\n\tat org.apache.spark.serializer.DeserializationStream$$anon$1.getNext(Serializer.scala:168)\r\n\tat org.apache.spark.util.NextIterator.hasNext(NextIterator.scala:73)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIterator(MemoryStore.scala:221)\r\n\tat org.apache.spark.storage.memory.MemoryStore.putIteratorAsValues(MemoryStore.scala:299)\r\n\tat org.apache.spark.storage.BlockManager.maybeCacheDiskValuesInMemory(BlockManager.scala:1516)\r\n\tat org.apache.spark.storage.BlockManager.getLocalValues(BlockManager.scala:825)\r\n\tat org.apache.spark.storage.BlockManager.get(BlockManager.scala:1111)\r\n\tat org.apache.spark.storage.BlockManager.getOrElseUpdate(BlockManager.scala:1178)\r\n\tat org.apache.spark.rdd.RDD.getOrCompute(RDD.scala:360)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:311)\r\n\tat org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)\r\n\tat org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:349)\r\n\tat org.apache.spark.rdd.RDD.iterator(RDD.scala:313)\r\n\tat org.apache.spark.shuffle.ShuffleWriteProcessor.write(ShuffleWriteProcessor.scala:59)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:99)\r\n\tat org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:52)\r\n\tat org.apache.spark.scheduler.Task.run(Task.scala:127)\r\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ERROR:root:Exception while sending command.\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1181, in send_command\n",
      "    answer = smart_decode(self.stream.readline()[:-1])\n",
      "  File \"C:\\Users\\hites\\Anaconda3\\lib\\socket.py\", line 589, in readinto\n",
      "    return self._sock.recv_into(b)\n",
      "ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1014, in send_command\n",
      "    response = connection.send_command(command)\n",
      "  File \"C:\\Spark_installed\\spark-3.0.0-preview2-bin-hadoop2.7\\python\\lib\\py4j-0.10.8.1-src.zip\\py4j\\java_gateway.py\", line 1193, in send_command\n",
      "    \"Error while receiving\", e, proto.ERROR_ON_RECEIVE)\n",
      "py4j.protocol.Py4JNetworkError: Error while receiving\n"
     ]
    }
   ],
   "source": [
    "# 5-fold Cross Validator Pipeline and fitting the train data using this cv pipeline\n",
    "\n",
    "cv_rf = CrossValidator(estimator=rfModel_new, estimatorParamMaps=paramGrid_rft, evaluator=evaluator_rfb, numFolds=5,seed=42).fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Predicting the test data using fitted cv pipeline\n",
    "\n",
    "pred_rft = cv_rf.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.7819275566312943\n"
     ]
    }
   ],
   "source": [
    "# AUC Score of tuned RF model\n",
    "\n",
    "print(\"AUC Score is\", evaluator_rfb.evaluate(pred_rft))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 216,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='RandomForestClassifier_8713e549bc1f', name='featuresCol', doc='features column name.'): 'features',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='labelCol', doc='label column name.'): 'label',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='predictionCol', doc='prediction column name.'): 'prediction',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities.'): 'probability',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name.'): 'rawPrediction',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='seed', doc='random seed.'): 42,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='cacheNodeIds', doc='If false, the algorithm will pass trees to executors to match instances with nodes. If true, the algorithm will cache node IDs for each instance. Caching can speed up training of deeper trees. Users can set how often should the cache be checkpointed or disable it by setting checkpointInterval.'): False,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='checkpointInterval', doc='set checkpoint interval (>= 1) or disable checkpoint (-1). E.g. 10 means that the cache will get checkpointed every 10 iterations. Note: this setting will be ignored if the checkpoint directory is not set in the SparkContext.'): 10,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='featureSubsetStrategy', doc=\"The number of features to consider for splits at each tree node. Supported options: 'auto' (choose automatically for task: If numTrees == 1, set to 'all'. If numTrees > 1 (forest), set to 'sqrt' for classification and to 'onethird' for regression), 'all' (use all features), 'onethird' (use 1/3 of the features), 'sqrt' (use sqrt(number of features)), 'log2' (use log2(number of features)), 'n' (when n is in the range (0, 1.0], use n * number of features. When n is in the range (1, number of features), use n features). default = 'auto'\"): 'auto',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='impurity', doc='Criterion used for information gain calculation (case-insensitive). Supported options: entropy, gini'): 'entropy',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='leafCol', doc='Leaf indices column name. Predicted leaf index of each instance in each tree by preorder.'): '',\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='maxBins', doc='Max number of bins for discretizing continuous features.  Must be >=2 and >= number of categories for any categorical feature.'): 32,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='maxDepth', doc='Maximum depth of the tree. (>= 0) E.g., depth 0 means 1 leaf node; depth 1 means 1 internal node + 2 leaf nodes.'): 10,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='maxMemoryInMB', doc='Maximum memory in MB allocated to histogram aggregation. If too small, then 1 node will be split per iteration, and its aggregates may exceed this size.'): 256,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='minInfoGain', doc='Minimum information gain for a split to be considered at a tree node.'): 0.0,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='minInstancesPerNode', doc='Minimum number of instances each child must have after split. If a split causes the left or right child to have fewer than minInstancesPerNode, the split will be discarded as invalid. Should be >= 1.'): 1,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='minWeightFractionPerNode', doc='Minimum fraction of the weighted sample count that each child must have after split. If a split causes the fraction of the total weight in the left or right child to be less than minWeightFractionPerNode, the split will be discarded as invalid. Should be in interval [0.0, 0.5).'): 0.0,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='numTrees', doc='Number of trees to train (>= 1).'): 60,\n",
       " Param(parent='RandomForestClassifier_8713e549bc1f', name='subsamplingRate', doc='Fraction of the training data used for learning each decision tree, in range (0, 1].'): 1.0}"
      ]
     },
     "execution_count": 216,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Best Model Hyper Parameters after tuning\n",
    "\n",
    "cv_rf.bestModel.stages[-1].extractParamMap()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "60"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Best model number of trees parameter from Grid Search\n",
    "\n",
    "cv_rf.bestModel.stages[-1].getNumTrees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6788441955193483"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Accuracy of the model on test data\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]\n",
    "\n",
    "binary_prediction=pred_rft.select(\"prediction\").collect()\n",
    "\n",
    "binary_true_labels=us_test_cat.select(\"Severity\").collect()\n",
    "\n",
    "np.sum(list([int(binary_true_labels[i][0]==binary_prediction[i][0]) for i in range(len(true_labels))]))/len(true_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_rft=pred_rft.toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.61      0.72    131790\n",
      "           1       0.51      0.81      0.62     64610\n",
      "\n",
      "    accuracy                           0.68    196400\n",
      "   macro avg       0.69      0.71      0.67    196400\n",
      "weighted avg       0.75      0.68      0.69    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_rft,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of RF Grid Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_rfg = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], cv_rf.bestModel.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from Random Forest Grid')"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of RF Grid Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_rfg['column'][:10], y=feat_imp_tuned_rfg['weight'][:10],data=feat_imp_tuned_rfg)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from Random Forest Grid\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# GBT Base Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize the GBT Base model\n",
    "\n",
    "gbt = GBTClassifier(seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Pipeline with stages to be used to fit the train data\n",
    "\n",
    "gbt_pipe = Pipeline(stages=[label_stringIdx, va, gbt])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fitting the training data using the pipeline above\n",
    "\n",
    "gbtModel = gbt_pipe.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.7842374103722184\n"
     ]
    }
   ],
   "source": [
    "# AUC Score from the model on the test data\n",
    "\n",
    "print(\"AUC Score is\", evaluator_rfb.evaluate(gbtModel.transform(us_test_cat)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_gbtn=(gbtModel.transform(us_test_cat)).toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True labels from the test data for the target variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.60      0.72    131790\n",
      "           1       0.51      0.83      0.63     64610\n",
      "\n",
      "    accuracy                           0.68    196400\n",
      "   macro avg       0.69      0.72      0.67    196400\n",
      "weighted avg       0.75      0.68      0.69    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_gbtn,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of GBT Base Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_gtbb = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], gbtModel.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from GBT Base Model')"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of GBT Base Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_gtbb['column'][:10], y=feat_imp_tuned_gtbb['weight'][:10],data=feat_imp_tuned_gtbb)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from GBT Base Model\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# GBT Binary Classification Grid Search"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# GBT Binary Tuned Best Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Initializing GBT Grid Pipeline \n",
    "\n",
    "gbt_t_new = GBTClassifier(maxIter=55,seed=42)\n",
    "\n",
    "# Creating pipeline for GBT grid Model \n",
    "\n",
    "gbt_pipe_t_new = Pipeline(stages=[label_stringIdx, va, gbt_t_new])\n",
    "\n",
    "# Binary Classification Evaluator\n",
    "\n",
    "evaluator = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')\n",
    "\n",
    "# Creating Grid Search for Hyper Parameter Tuning for GBT model\n",
    "\n",
    "grid_gbt_t_new = ParamGridBuilder().addGrid(gbt_t_new.stepSize, [0.2,0.4,0.01]).addGrid(gbt_t_new.maxDepth, [3, 5, 8]).build()\n",
    "\n",
    "# Cross Validator Pipeline with 5 fold cv to fit the training data\n",
    "\n",
    "cv1_gbt_t_new = CrossValidator(estimator=gbt_pipe_t_new,estimatorParamMaps=grid_gbt_t_new, numFolds=5, evaluator=evaluator,seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fitting train data using 5-fold cross validator pipeline\n",
    "\n",
    "cvModel_gbt_t_new = cv1_gbt_t_new.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.8118245222266022\n"
     ]
    }
   ],
   "source": [
    "# AUC Score from the fitted pipeline for the test data\n",
    "\n",
    "print(\"AUC Score is\", evaluator.evaluate(cvModel_gbt_t_new.transform(us_test_cat)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7997203339021297"
      ]
     },
     "execution_count": 173,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# AUC Score from the Cross validator Pipeline\n",
    "\n",
    "np.max(cvModel_gbt_t_new.avgMetrics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='GBTClassifier_48357a426a79', name='featuresCol', doc='features column name.'): 'features',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='labelCol', doc='label column name.'): 'label',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='predictionCol', doc='prediction column name.'): 'prediction',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities.'): 'probability',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name.'): 'rawPrediction',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='seed', doc='random seed.'): 42,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='cacheNodeIds', doc='If false, the algorithm will pass trees to executors to match instances with nodes. If true, the algorithm will cache node IDs for each instance. Caching can speed up training of deeper trees. Users can set how often should the cache be checkpointed or disable it by setting checkpointInterval.'): False,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='checkpointInterval', doc='set checkpoint interval (>= 1) or disable checkpoint (-1). E.g. 10 means that the cache will get checkpointed every 10 iterations. Note: this setting will be ignored if the checkpoint directory is not set in the SparkContext.'): 10,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='featureSubsetStrategy', doc=\"The number of features to consider for splits at each tree node. Supported options: 'auto' (choose automatically for task: If numTrees == 1, set to 'all'. If numTrees > 1 (forest), set to 'sqrt' for classification and to 'onethird' for regression), 'all' (use all features), 'onethird' (use 1/3 of the features), 'sqrt' (use sqrt(number of features)), 'log2' (use log2(number of features)), 'n' (when n is in the range (0, 1.0], use n * number of features. When n is in the range (1, number of features), use n features). default = 'auto'\"): 'all',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='impurity', doc='Criterion used for information gain calculation (case-insensitive). Supported options: variance'): 'variance',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='leafCol', doc='Leaf indices column name. Predicted leaf index of each instance in each tree by preorder.'): '',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='lossType', doc='Loss function which GBT tries to minimize (case-insensitive). Supported options: logistic'): 'logistic',\n",
       " Param(parent='GBTClassifier_48357a426a79', name='maxBins', doc='Max number of bins for discretizing continuous features.  Must be >=2 and >= number of categories for any categorical feature.'): 32,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='maxDepth', doc='Maximum depth of the tree. (>= 0) E.g., depth 0 means 1 leaf node; depth 1 means 1 internal node + 2 leaf nodes.'): 8,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='maxIter', doc='max number of iterations (>= 0).'): 55,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='maxMemoryInMB', doc='Maximum memory in MB allocated to histogram aggregation. If too small, then 1 node will be split per iteration, and its aggregates may exceed this size.'): 256,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='minInfoGain', doc='Minimum information gain for a split to be considered at a tree node.'): 0.0,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='minInstancesPerNode', doc='Minimum number of instances each child must have after split. If a split causes the left or right child to have fewer than minInstancesPerNode, the split will be discarded as invalid. Should be >= 1.'): 1,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='minWeightFractionPerNode', doc='Minimum fraction of the weighted sample count that each child must have after split. If a split causes the fraction of the total weight in the left or right child to be less than minWeightFractionPerNode, the split will be discarded as invalid. Should be in interval [0.0, 0.5).'): 0.0,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='stepSize', doc='Step size (a.k.a. learning rate) in interval (0, 1] for shrinking the contribution of each estimator.'): 0.2,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='subsamplingRate', doc='Fraction of the training data used for learning each decision tree, in range (0, 1].'): 1.0,\n",
       " Param(parent='GBTClassifier_48357a426a79', name='validationTol', doc='Threshold for stopping early when fit with validation is used. If the error rate on the validation input changes by less than the validationTol, then learning will stop early (before `maxIter`). This parameter is ignored when fit without validation is used.'): 0.01}"
      ]
     },
     "execution_count": 217,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Hyper pramaters from the best model \n",
    "\n",
    "cvModel_gbt_t_new.bestModel.stages[-1].extractParamMap()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_gbt_t_new=cvModel_gbt_t_new.transform(us_test_cat).toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.65      0.75    131790\n",
      "           1       0.53      0.82      0.65     64610\n",
      "\n",
      "    accuracy                           0.70    196400\n",
      "   macro avg       0.71      0.73      0.70    196400\n",
      "weighted avg       0.77      0.70      0.71    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_gbt_t_new,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of GBT Grid Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_gbt_t_new = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], cvModel_gbt_t_new.bestModel.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from GBT Best tuned')"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of GBT Grid Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_gbt_t_new['column'][:10], y=feat_imp_tuned_gbt_t_new['weight'][:10],data=feat_imp_tuned_gbt_t_new)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from GBT Best tuned\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Decision Trees Binary Classification Base Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Create initial Decision Tree Model\n",
    "dt = DecisionTreeClassifier(labelCol=\"label\", featuresCol=\"features\",seed=42)\n",
    "\n",
    "# Pipeline with stages created for DT Model \n",
    "\n",
    "dt_pipe = Pipeline(stages=[label_stringIdx, va, dt])\n",
    "\n",
    "# Train model with Training Data\n",
    "\n",
    "dtModel = dt_pipe.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Binary Class Evaluator Initialize\n",
    "\n",
    "evaluator = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Transform the test data to get prediction from the model for the test data\n",
    "\n",
    "pred_dt = dtModel.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.7039599188458128\n"
     ]
    }
   ],
   "source": [
    "# AUC Score from the evaluator for the test data\n",
    "\n",
    "print(\"AUC Score is\",evaluator.evaluate(pred_dt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6643584521384929"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Accuracy Calculation for DT Base Model\n",
    "\n",
    "binary_prediction=pred_dt.select(\"prediction\").collect()\n",
    "\n",
    "binary_true_labels=us_test_cat.select(\"Severity\").collect()\n",
    "\n",
    "np.sum(list([int(binary_true_labels[i][0]==binary_prediction[i][0]) for i in range(len(true_labels))]))/len(true_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_dtb=pred_dt.toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.86      0.60      0.70    131790\n",
      "           1       0.49      0.81      0.61     64610\n",
      "\n",
      "    accuracy                           0.66    196400\n",
      "   macro avg       0.68      0.70      0.66    196400\n",
      "weighted avg       0.74      0.66      0.67    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_dtb,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of DT Base Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_dtb= pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], dtModel.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from DT Binary Base Model')"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of DT Base Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_dtb['column'][:10], y=feat_imp_tuned_dtb['weight'][:10],data=feat_imp_tuned_dtb)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from DT Binary Base Model\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Decision Tree Binary Classification Grid Search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing DT Grid Pipeline\n",
    "\n",
    "dt_new = DecisionTreeClassifier(labelCol=\"label\", featuresCol=\"features\",seed=42)\n",
    "\n",
    "# Creating pipeline for DT Grid Model \n",
    "\n",
    "dt_new_pipe = Pipeline(stages=[label_stringIdx, va, dt_new])\n",
    "\n",
    "# Binary Evaluator Initializing\n",
    "\n",
    "evaluator = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')\n",
    "\n",
    "# Creating Grid Search for Hyper Parameter Tuning for DT Model\n",
    "\n",
    "grid_dt = ParamGridBuilder().addGrid(dt_new.maxDepth, [10,15,30]).addGrid(dt_new.minInstancesPerNode, [500,1000,1500]).addGrid(dt_new.maxBins,[20,35,50]).build()\n",
    "\n",
    "# Cross Validator Pipeline with 5 fold cv to fit the training data\n",
    "\n",
    "cv1_dt = CrossValidator(estimator=dt_new_pipe,estimatorParamMaps=grid_dt, numFolds=5, evaluator=evaluator,seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fitting the train data using the 5-fold Cross validator pipeline\n",
    "\n",
    "dtModel_t = cv1_dt.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Predicting the test data using the fitted pipeline\n",
    "\n",
    "pred_dtt = dtModel_t.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score 0.6710458592255817\n"
     ]
    }
   ],
   "source": [
    "# AUC Score for the fitted pipeline for test data\n",
    "\n",
    "print(\"AUC Score\", evaluator.evaluate(pred_dtt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='featuresCol', doc='features column name.'): 'features',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='labelCol', doc='label column name.'): 'label',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='predictionCol', doc='prediction column name.'): 'prediction',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities.'): 'probability',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name.'): 'rawPrediction',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='seed', doc='random seed.'): 42,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='cacheNodeIds', doc='If false, the algorithm will pass trees to executors to match instances with nodes. If true, the algorithm will cache node IDs for each instance. Caching can speed up training of deeper trees. Users can set how often should the cache be checkpointed or disable it by setting checkpointInterval.'): False,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='checkpointInterval', doc='set checkpoint interval (>= 1) or disable checkpoint (-1). E.g. 10 means that the cache will get checkpointed every 10 iterations. Note: this setting will be ignored if the checkpoint directory is not set in the SparkContext.'): 10,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='impurity', doc='Criterion used for information gain calculation (case-insensitive). Supported options: entropy, gini'): 'gini',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='leafCol', doc='Leaf indices column name. Predicted leaf index of each instance in each tree by preorder.'): '',\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='maxBins', doc='Max number of bins for discretizing continuous features.  Must be >=2 and >= number of categories for any categorical feature.'): 50,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='maxDepth', doc='Maximum depth of the tree. (>= 0) E.g., depth 0 means 1 leaf node; depth 1 means 1 internal node + 2 leaf nodes.'): 10,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='maxMemoryInMB', doc='Maximum memory in MB allocated to histogram aggregation. If too small, then 1 node will be split per iteration, and its aggregates may exceed this size.'): 256,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='minInfoGain', doc='Minimum information gain for a split to be considered at a tree node.'): 0.0,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='minInstancesPerNode', doc='Minimum number of instances each child must have after split. If a split causes the left or right child to have fewer than minInstancesPerNode, the split will be discarded as invalid. Should be >= 1.'): 1500,\n",
       " Param(parent='DecisionTreeClassifier_0bccbe6ff0a3', name='minWeightFractionPerNode', doc='Minimum fraction of the weighted sample count that each child must have after split. If a split causes the fraction of the total weight in the left or right child to be less than minWeightFractionPerNode, the split will be discarded as invalid. Should be in interval [0.0, 0.5).'): 0.0}"
      ]
     },
     "execution_count": 218,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Best Model Hyper parameters for the DT Grid Search Model\n",
    "\n",
    "dtModel_t.bestModel.stages[-1].extractParamMap()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_dtbt=pred_dtt.toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.57      0.70    131790\n",
      "           1       0.49      0.85      0.62     64610\n",
      "\n",
      "    accuracy                           0.66    196400\n",
      "   macro avg       0.69      0.71      0.66    196400\n",
      "weighted avg       0.76      0.66      0.67    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_dtbt,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating Pandas Dataframe for Features and their Importance of DT Grid Model for Binary Classification\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_dtbt= pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], dtModel_t.bestModel.stages[-1].featureImportances)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Top 10 Features based on Importance from DT Binary Grid Model')"
      ]
     },
     "execution_count": 172,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting top 10 Features from Feature Importance of DT Grid Model for Binary Classification\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "sns.barplot(x=feat_imp_tuned_dtbt['column'][:10], y=feat_imp_tuned_dtbt['weight'][:10],data=feat_imp_tuned_dtbt)\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Features\")\n",
    "plt.ylabel(\"Weights\")\n",
    "plt.title(\"Top 10 Features based on Importance from DT Binary Grid Model\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Regression "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Base Model Binary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Standard Scaler to standardize the output of vector assembler before feeding it to Logistic Regression\n",
    "\n",
    "center = feature.StandardScaler(withMean=True, withStd=False, inputCol='features', outputCol='centered_features')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create initial LogisticRegression model\n",
    "lr = LogisticRegression(labelCol=\"label\", featuresCol=\"centered_features\")\n",
    "\n",
    "# Pipeline for training data \n",
    "\n",
    "lrModel = Pipeline(stages=[label_stringIdx,va, center, lr])\n",
    "\n",
    "# Fit the train data using LR model\n",
    "\n",
    "lr_fit = lrModel.fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Transsform test data to predict Severity by using fitted pipeline from training data\n",
    "\n",
    "pred_lrb = lr_fit.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Area Under ROC for Logistic Base Model  0.7619402091983631\n"
     ]
    }
   ],
   "source": [
    "# Evaluator to get AUC Score for test data\n",
    "\n",
    "evaluator_lrb = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')\n",
    "print('Test Area Under ROC for Logistic Base Model ', evaluator_lrb.evaluate(lr_fit.transform(us_test_cat)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set areaUnderROC: 0.7627619042131226\n"
     ]
    }
   ],
   "source": [
    "# ROC Curve for LR Base Model\n",
    "\n",
    "trainingSummary = lr_fit.stages[-1].summary\n",
    "roc = trainingSummary.roc.toPandas()\n",
    "plt.plot(roc['FPR'],roc['TPR'])\n",
    "plt.ylabel('False Positive Rate')\n",
    "plt.xlabel('True Positive Rate')\n",
    "plt.title('ROC Curve')\n",
    "plt.grid(True)\n",
    "plt.show()\n",
    "print('Training set areaUnderROC: ' + str(trainingSummary.areaUnderROC))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Logistic Regression Coefficients to Array \n",
    "\n",
    "coef_l1=lr_fit.stages[-1].coefficientMatrix.toArray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "# to stack the coefficient array column wise for further analysis\n",
    "\n",
    "cof_l1=np.hstack(coef_l1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating pandas dataframe with Logistic Regression weights for each variable along with variable name\n",
    "\n",
    "pd.set_option('display.max_rows', None)\n",
    "feat_imp_tuned_b = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], cof_l1)),\n",
    "            columns = ['column', 'weight']).sort_values('weight',ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Coefficient from LR model for each variable\n",
    "\n",
    "coef_L1=lr_fit.stages[-1].coefficients.toArray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 6.10404226e-02,  3.84503967e-01,  5.55462923e-01,  2.83682423e-02,\n",
       "        3.11545767e-01,  5.72957889e-01,  2.68684159e-01,  6.40513508e-01,\n",
       "        2.39662302e-01,  7.29093517e-01,  3.38870399e-01,  6.35365038e-01,\n",
       "        3.42524554e-01,  3.00627117e-01,  2.86496247e-01,  3.33305131e-01,\n",
       "        1.74724528e-01,  7.28250150e-01,  4.11683887e-01,  1.01770132e-01,\n",
       "        7.77776525e-02,  4.80918989e-01,  1.93834032e-01,  6.10964689e-02,\n",
       "        7.07937460e-02, -2.46781145e-02,  3.25562566e-02,  1.93521868e+00,\n",
       "        6.98425343e-01,  2.44544359e-01, -2.92797816e-01,  4.20970924e-01,\n",
       "        1.18834527e+00, -4.98702447e-01,  1.49876162e+00, -3.76084559e-01,\n",
       "        8.73144768e-01,  1.01459587e+00, -3.68386095e-01,  8.15470313e-01,\n",
       "        1.02401163e+00,  1.33550551e+00,  9.34355323e-01, -5.02078892e-02,\n",
       "        1.90952537e+00,  1.15563651e+00,  6.80595337e-01,  5.58708379e-02,\n",
       "        2.25339626e-02,  2.20958954e-02,  7.43138848e-04,  6.98986882e-01,\n",
       "        6.92230035e-01,  3.43491620e-01,  1.09905601e-01, -6.59910295e-02,\n",
       "        2.73193089e-01,  6.57264786e-02,  1.43917276e-01,  6.74550394e-05,\n",
       "        7.00180998e-02,  2.05815870e-01,  3.62644772e-01,  2.44123017e-01,\n",
       "        8.74585264e-02,  2.65865377e-02,  1.39148506e-02, -3.64614378e-02,\n",
       "        3.60980672e-02,  5.39018526e-02,  6.97262794e-02,  3.48474891e-02,\n",
       "        7.32900545e-02,  9.03299822e-03,  3.43603153e-02,  1.55938700e-02,\n",
       "        1.39685633e-02,  3.34354498e-02,  4.88498457e-02,  7.42683555e-02,\n",
       "        6.41895294e-03, -1.52099373e+00, -2.92797816e-01,  3.21061922e-01,\n",
       "        1.12504363e-01,  9.90974565e-02,  8.69937407e-01, -3.11252258e-01,\n",
       "       -4.64919144e-02,  7.05797649e-02,  2.02713308e-01, -2.74288756e-03,\n",
       "        4.03122329e-01,  2.52918849e-01,  3.42967380e-01, -3.63234523e-01,\n",
       "       -1.65340324e-01,  2.05437823e-01,  2.01133559e-01,  3.13500571e-01,\n",
       "        3.03022094e-01, -3.83431591e-03,  1.29392900e-04,  5.04152924e-02,\n",
       "        1.42110421e-02,  9.75298836e-03,  3.16315594e-02, -7.36661603e-01,\n",
       "        3.93267264e-01, -6.70645537e-01,  3.50216355e-01,  1.91966480e-01,\n",
       "        3.17809826e-02,  1.69946149e-01, -2.11351631e+00, -4.39883557e-01,\n",
       "       -1.60607312e+00, -9.26287099e-01, -1.28426974e+00])"
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_L1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of features are 119\n",
      "Eliminated features out of 119 are 0\n"
     ]
    }
   ],
   "source": [
    "# Taking absolute values of weights and calculating the number of features eliminated by LR Model after L1 regularization \n",
    "\n",
    "coef_L1 = np.absolute(coef_L1)\n",
    "\n",
    "print('Total number of features are',len(coef_L1))\n",
    "\n",
    "sorted_abs = np.sort(coef_L1)\n",
    "\n",
    "weights_notzero = sorted_abs[sorted_abs == 0]\n",
    "nonzero_weights = len(sorted_abs[sorted_abs == 0])\n",
    "\n",
    "print('Eliminated features out of ' + str(len(coef_L1)) + ' are', nonzero_weights)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_lrbal=pred_lrb.toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.85      0.63      0.72    131790\n",
      "           1       0.50      0.77      0.61     64610\n",
      "\n",
      "    accuracy                           0.67    196400\n",
      "   macro avg       0.67      0.70      0.66    196400\n",
      "weighted avg       0.73      0.67      0.68    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_lrbal,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Binary Grid Search Binary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Logistic Regression Pipeline initialization\n",
    "\n",
    "lr_new = LogisticRegression(labelCol=\"label\", featuresCol=\"centered_features\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Grid Search for tuning the hyper parameters of Logistic Regression Model\n",
    "\n",
    "paramGrid_lr = ParamGridBuilder().addGrid(lr_new.regParam, [0.01, 0.04,0.07]).addGrid(lr_new.elasticNetParam, [0.1,0.4,0.7]).build()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating pipeline to be used for fitting the training data\n",
    "\n",
    "cvModel_lrbal = Pipeline(stages=[label_stringIdx,va,center,lr_new])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Binary Evaluator for evaluating the model performance\n",
    "\n",
    "evaluator_lrbt = BinaryClassificationEvaluator(labelCol='label',metricName='areaUnderROC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Cross validator pipeline initialization for 5-fold cross validation and fitting the train data\n",
    "\n",
    "cv_lrbal = CrossValidator(estimator=cvModel_lrbal, estimatorParamMaps=paramGrid_lr, evaluator=evaluator_lrbt, numFolds=5,seed=42).fit(us_train_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Testing the test data on fitted 5 fold cv pipeline\n",
    "\n",
    "pred_lrbalt=cv_lrbal.transform(us_test_cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC Score is 0.7612376191461516\n"
     ]
    }
   ],
   "source": [
    "# Accuracy calculation from the Multiclass evaluator\n",
    "\n",
    "print(\"AUC Score is\",evaluator_lrbt.evaluate(pred_lrbalt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set areaUnderROC: 0.7622374995379829\n"
     ]
    }
   ],
   "source": [
    "# ROC Curve for Logistic Regression Grid Search Model\n",
    "\n",
    "trainingSummary_t = cv_lrbal.bestModel.stages[-1].summary\n",
    "roc = trainingSummary_t.roc.toPandas()\n",
    "plt.plot(roc['FPR'],roc['TPR'])\n",
    "plt.ylabel('False Positive Rate')\n",
    "plt.xlabel('True Positive Rate')\n",
    "plt.title('ROC Curve')\n",
    "plt.grid(True)\n",
    "plt.show()\n",
    "print('Training set areaUnderROC: ' + str(trainingSummary_t.areaUnderROC))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{Param(parent='LogisticRegression_f000cca535b6', name='aggregationDepth', doc='suggested depth for treeAggregate (>= 2).'): 2,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='elasticNetParam', doc='the ElasticNet mixing parameter, in range [0, 1]. For alpha = 0, the penalty is an L2 penalty. For alpha = 1, it is an L1 penalty.'): 0.1,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='featuresCol', doc='features column name.'): 'centered_features',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='fitIntercept', doc='whether to fit an intercept term.'): True,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='labelCol', doc='label column name.'): 'label',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='predictionCol', doc='prediction column name.'): 'prediction',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='probabilityCol', doc='Column name for predicted class conditional probabilities. Note: Not all models output well-calibrated probability estimates! These probabilities should be treated as confidences, not precise probabilities.'): 'probability',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='rawPredictionCol', doc='raw prediction (a.k.a. confidence) column name.'): 'rawPrediction',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='standardization', doc='whether to standardize the training features before fitting the model.'): True,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='threshold', doc='Threshold in binary classification prediction, in range [0, 1]. If threshold and thresholds are both set, they must match.e.g. if threshold is p, then thresholds must be equal to [1-p, p].'): 0.5,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='family', doc='The name of family which is a description of the label distribution to be used in the model. Supported options: auto, binomial, multinomial'): 'auto',\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='maxIter', doc='max number of iterations (>= 0).'): 100,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='regParam', doc='regularization parameter (>= 0).'): 0.01,\n",
       " Param(parent='LogisticRegression_f000cca535b6', name='tol', doc='the convergence tolerance for iterative algorithms (>= 0).'): 1e-06}"
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Best Model Hyper parameters after Tuning LR Model\n",
    "\n",
    "cv_lrbal.bestModel.stages[-1].extractParamMap()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6688034623217922"
      ]
     },
     "execution_count": 194,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Accuracy Calculation for the test data from the model\n",
    "\n",
    "binary_prediction=pred_lrbalt.select(\"prediction\").collect()\n",
    "\n",
    "binary_true_labels=us_test_cat.select(\"Severity\").collect()\n",
    "\n",
    "np.sum(list([int(binary_true_labels[i][0]==binary_prediction[i][0]) for i in range(len(true_labels))]))/len(true_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prediction output from the model to pandas\n",
    "\n",
    "prediction_lrbalt=pred_lrbalt.toPandas()[\"prediction\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [],
   "source": [
    "# True Labels from test data for Target Variable\n",
    "\n",
    "true_labels=us_test_cat.toPandas()[\"Severity\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initializing Classification Report from sklearn\n",
    "\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.85      0.61      0.71    131790\n",
      "           1       0.50      0.78      0.61     64610\n",
      "\n",
      "    accuracy                           0.67    196400\n",
      "   macro avg       0.67      0.70      0.66    196400\n",
      "weighted avg       0.74      0.67      0.68    196400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report Generation for all metrics display at once\n",
    "\n",
    "print(classification_report(y_pred=prediction_dtbalt,y_true=true_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Coefficient from LR model for each variable\n",
    "\n",
    "coef_L1_m=cv_lrbal.bestModel.stages[-1].coefficients.toArray()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.05730593,  0.1631614 ,  0.22683363, -0.0922915 ,  0.12743824,\n",
       "        0.22850115,  0.0874906 ,  0.29489116,  0.        ,  0.37672293,\n",
       "        0.15445802,  0.28753523,  0.14824354,  0.00970946,  0.10372804,\n",
       "        0.08653566,  0.        ,  0.37548958,  0.13350159, -0.01698079,\n",
       "       -0.04548731,  0.17180723,  0.01903713,  0.10730425,  0.08693924,\n",
       "        0.01656798,  0.06626178,  1.65740542,  0.5776856 ,  0.        ,\n",
       "       -0.25852766,  0.34445895,  0.99689203, -0.21432981,  1.33791048,\n",
       "        0.        ,  0.80841273,  0.59158904, -0.23910861,  0.58651763,\n",
       "        0.90778268,  0.11721702,  0.81946075,  0.        ,  1.58545586,\n",
       "        0.78832342,  0.50243283,  0.0249303 ,  0.        ,  0.        ,\n",
       "       -0.01444565,  0.63605211,  0.62860006,  0.22391409,  0.02995857,\n",
       "       -0.11490668,  0.14692048,  0.        ,  0.02469895, -0.06192017,\n",
       "        0.        ,  0.10814645,  0.23145346,  0.13481408,  0.03663532,\n",
       "        0.        ,  0.        , -0.03897284,  0.        ,  0.        ,\n",
       "        0.02532982,  0.00652907,  0.01371828,  0.        ,  0.        ,\n",
       "        0.        ,  0.        ,  0.        ,  0.01066444,  0.02208044,\n",
       "        0.        , -1.41648022, -0.25852766,  0.26103542,  0.        ,\n",
       "        0.        ,  0.        , -0.22986335,  0.0026555 ,  0.04903758,\n",
       "        0.14350738,  0.        ,  0.21757303,  0.08485045,  0.0949972 ,\n",
       "       -0.44105302,  0.        ,  0.01732823,  0.07618706,  0.        ,\n",
       "        0.23152556, -0.00195721,  0.        ,  0.0395699 ,  0.00836024,\n",
       "        0.01054067,  0.00280582, -0.64456   ,  0.        , -0.63672179,\n",
       "        0.17710428,  0.18114564,  0.        ,  0.0416721 , -0.79625681,\n",
       "       -0.37949686, -1.40585049, -0.38560954, -1.20185474])"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_L1_m"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Pandas dataframe of weights of variables with variable names to find which variables are eliminated\n",
    "\n",
    "feat_imp_tuned_lrt = pd.DataFrame(list(zip([i for i in us_train_cat.columns if i!='Severity'], np.absolute(coef_L1_m))),\n",
    "            columns = ['column', 'weight']).sort_values('weight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>column</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>80</td>\n",
       "      <td>Wind_Direction_Index_12</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>65</td>\n",
       "      <td>Wind_Direction_Index_7</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>85</td>\n",
       "      <td>clear</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>84</td>\n",
       "      <td>cloud</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>57</td>\n",
       "      <td>month_of_year_Index_11</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>TMC_Index_18</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>77</td>\n",
       "      <td>Wind_Direction_Index_5</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>76</td>\n",
       "      <td>Wind_Direction_Index_15</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75</td>\n",
       "      <td>Wind_Direction_Index_6</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>74</td>\n",
       "      <td>Wind_Direction_Index_13</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     column  weight\n",
       "80  Wind_Direction_Index_12     0.0\n",
       "65   Wind_Direction_Index_7     0.0\n",
       "85                    clear     0.0\n",
       "84                    cloud     0.0\n",
       "57   month_of_year_Index_11     0.0\n",
       "29             TMC_Index_18     0.0\n",
       "77   Wind_Direction_Index_5     0.0\n",
       "76  Wind_Direction_Index_15     0.0\n",
       "75   Wind_Direction_Index_6     0.0\n",
       "74  Wind_Direction_Index_13     0.0"
      ]
     },
     "execution_count": 211,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Sample of 10 features eliminated by the Logistic Regression Model after L1 Regularization\n",
    "\n",
    "feat_imp_tuned_lrt[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of features are 119\n",
      "Eliminated features out of 119 are 28\n"
     ]
    }
   ],
   "source": [
    "# Taking absolute values of weights and calculating the number of features eliminated by LR Model after L1 regularization \n",
    "\n",
    "coef_L1_m = np.absolute(coef_L1_m)\n",
    "\n",
    "print('Total number of features are',len(coef_L1_m))\n",
    "\n",
    "sorted_abs = np.sort(coef_L1_m)\n",
    "\n",
    "weights_notzero = sorted_abs[sorted_abs == 0]\n",
    "nonzero_weights = len(sorted_abs[sorted_abs == 0])\n",
    "\n",
    "print('Eliminated features out of ' + str(len(coef_L1_m)) + ' are', len(weights_notzero))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}