flight_status.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# What's My Flight Status?: Using Flight Data to Predict Flight Delays\n",
    "***\n",
    "Author: Andre Layton\n",
    "***\n",
    "\n",
    "## Overview\n",
    "\n",
    "One of the most common applications of supervised learning in the aviation industry is predicting flight delays. By analyzing historical data on flight delays, machine learning and deep learning algorithms can identify patterns and factors that correlate with delays. This information can then be used to predict if a particular flight is likely to be delayed. Carriers can use these predictions to take preemptive measures, such as adjusting schedules or re-routing passengers, in order to minimize the impact of delays and their sales.\n",
    "\n",
    "## Business Problem\n",
    "\n",
    "I’ve been hired to create an algorithm that can predict flight delays, which will eventually be deployed as an app for consumers to be able to track their flights. This is beneficial to both the airline and potential passengers – for the airline, it will help with flight logistics and reduce fees due to delays (i.e. tarmac fees, reimbursements, etc.). For passengers, the app will allow them to make delay arrangements and take measures ahead of time, and possibly save on delay expenses. While delays are frustrating whether expected or not, United aims to use this strategy to display company honesty and gain more control over their flights.\n",
    "\n",
    ">Note: The ultimate objective is to develop an app for consumer use, but within the constraints of this analysis, the model will be saved and stored in the repository [here.](/best_model.h5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"https://res.cloudinary.com/sagacity/image/upload/c_crop,h_2799,w_4199,x_0,y_0/c_limit,dpr_auto,f_auto,fl_lossy,q_80,w_1080/shutterstock_402702424_dlgrgk.jpg\" style=\"height:650px\" />"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Understanding \n",
    "\n",
    "To start, I import all the necessary packages, and I set a seed for reproducibility purposes. Then, I load the data, which is split into two files - a text file (`.txt`) and a CSV file (`.csv`). The text file contains the metadata, which in this case contains various column names and a short description. The CSV file contains the flight data needed for analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import relevant libraries\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "from matplotlib.ticker import AutoMinorLocator\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "from sklearn.pipeline import make_pipeline, Pipeline\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV, cross_validate\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from collections import Counter\n",
    "from sklearn.preprocessing import MinMaxScaler, OneHotEncoder\n",
    "from sklearn.metrics import precision_score, recall_score, f1_score, confusion_matrix, classification_report,\\\n",
    "                            ConfusionMatrixDisplay, make_scorer\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier, AdaBoostClassifier\n",
    "import joblib\n",
    "from tensorflow.keras.models import Sequential, load_model\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from tensorflow.keras import regularizers, optimizers\n",
    "\n",
    "# Create a seed for reproducibility\n",
    "seed=24"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AIRPORT_COORDINATES\n",
      "\tORIGIN_AIRPORT_ID: \tAirport ID, matches to ORIGIN_AIRPORT_ID in other files\n",
      "\tDISPLAY_AIRPORT_NAME:  \tDisplay Airport, matches to DISPLAY_AIRPORT_NAME in other files\n",
      "\tLATITUDE: \t\tLatitude for airport\n",
      "\tLONGITUDE: \t\tLongitude for airport\n",
      "\n",
      "B43_AIRCRAFT_INVENTORY\n",
      "\tMANUFACTURE_YEAR: \tManufacture year\n",
      "\tTAIL_NUM: \t\tUnique tail number, matches to TAIL_NUM in other files\n",
      "\tNUMBER_OF_SEATS: \tNumber of seats on aircraft\n",
      "\t\n",
      "CARRIER_DECODE\n",
      "\tAIRLINE_ID: \t\tAirport ID, matches to AIRLINE_ID in other files\n",
      "\tOP_UNIQUE_CARRIER: \tCarrier code, matches to OP_UNIQUE_CARRIER in other files\n",
      "\tCARRIER_NAME: \t\tCarrier name, matches to UNIQUE_CARRIER_NAME or CARRIER_NAME in other files\n",
      "\n",
      "ONTIME_REPORTING_XX\n",
      "\tMONTH: \t\t\tMonth\n",
      "\tDAY_OF_MONTH: \t\tDay of the month (1-31)\n",
      "\tDAY_OF_WEEK: \t\tDay of the week\n",
      "\tOP_UNIQUE_CARRIER: \tCarrier code, matches to OP_UNIQUE_CARRIER in other files\n",
      "\tTAIL_NUM: \t\tUnique tail number, matches to TAIL_NUM in other files\n",
      "\tOP_CARRIER_FL_NUM: \tFlight number\n",
      "\tORIGIN_AIRPORT_ID: \tAirport ID, matches to ORIGIN_AIRPORT_ID in other files\n",
      "\tORIGIN: \t\tOrigin airport abbreviation\n",
      "\tORIGIN_CITY_NAME: \tOrigin city name\n",
      "\tDEST_AIRPORT_ID: \tDestination airport ID, matches Airport ID in other files\n",
      "\tDEST: \t\t\tDestination airport abbreviation\n",
      "\tDEST_CITY_NAME: \tDestination city name\n",
      "\tCRS_DEP_TIME: \t\tPlanned departure time\n",
      "\tDEP_TIME: \t\tActual departure time\n",
      "\tDEP_DELAY_NEW: \t\tDeparture delay in minutes\n",
      "\tDEP_DEL15:\t\tTARGET VARIABLE Binary if delayed over 15 min, 1 is yes\n",
      "\tDEP_TIME_BLK:\t\tDeparture time block\n",
      "\tCRS_ARR_TIME:\t\tPlanned arrival time\n",
      "\tARR_TIME:\t\tActual arrival time\n",
      "\tARR_DELAY_NEW:\t\tArrival delay in minutes\n",
      "\tARR_TIME_BLK:\t\tArrival time block\n",
      "\tCANCELLED:\t\tFlag if flight was cancelled\n",
      "\tCANCELLATION_CODE:\tCancellation Code\n",
      "\tCRS_ELAPSED_TIME:\tFlight planned elapsed time\n",
      "\tACTUAL_ELAPSED_TIME:\tFlight actual elapsed time\n",
      "\tDISTANCE:\t\tFlight Distance in miles\n",
      "\tDISTANCE_GROUP:\t\tFlight distance group\n",
      "\tCARRIER_DELAY:\t\tFlag for a carrier delay\n",
      "\tWEATHER_DELAY:\t\tFlag for a weather delay\n",
      "\tNAS_DELAY:\t\tFlag for a NAS delay\n",
      "\tSECURITY_DELAY:\t\tFlag for a security delay\n",
      "\tLATE_AIRCRAFT_DELAY:\tFlag for a late aircraft delay\n",
      "\n",
      "P10_EMPLOYEES\n",
      "\tYEAR: \t\t\tYear\n",
      "\tAIRLINE_ID: \t\tAirport ID, matches to AIRLINE_ID in other files\n",
      "\tOP_UNIQUE_CARRIER: \tCarrier code, matches to OP_UNIQUE_CARRIER in other files\n",
      "\tUNIQUE_CARRIER_NAME: \tCarrier name, matches to UNIQUE_CARRIER_NAME in other files\n",
      "\tCARRIER: \t\tCarrier abbreviation\n",
      "\tCARRIER_NAME: \t\tCarrier name, matches to UNIQUE_CARRIER_NAME or CARRIER_NAME in other files\n",
      "\tENTITY:\t\t\t\n",
      "\tGENERAL_MANAGE:\t\tGeneral managers\n",
      "\tPILOTS_COPILOTS:\tPilots/Copilots\n",
      "\tOTHER_FLT_PERS:\t\tOther flight personnel\n",
      "\tPASS_GEN_SVC_ADMIN:\tPassenger/General Services & Administration\n",
      "\tMAINTENANCE:\t\tMaintenance Employees\n",
      "\tARCFT_TRAF_HANDLING_GRP1: Aircraft Traffic Handling Group1 Employees\n",
      "\tGEN_ARCFT_TRAF_HANDLING\t: General Aircraft Traffic Handling Employees\n",
      "\tAIRCRAFT_CONTROL:\tAircraft Control Employees\n",
      "\tPASSENGER_HANDLING:\tPassenger Handling Employees\n",
      "\tCARGO_HANDLING:\t\tCargo Handling Employees\n",
      "\tTRAINEES_INTRUCTOR:\tTrainees and Instructor\n",
      "\tSTATISTICAL:\t\tStatistical Employees\n",
      "\tTRAFFIC_SOLICITERS:\tTraffic Soliciters\n",
      "\tOTHER:\t\t\tOther Employees\n",
      "\tTRANSPORT_RELATED:\tTransport Related Employees\n",
      "\tTOTAL:\t\t\tTotal employees\n",
      "\n",
      "T3_AIR_CARRIER_SUMMARY_AIRPORT_ACTIVITY_XXXX\n",
      "\tOP_UNIQUE_CARRIER: \tCarrier code, matches to OP_UNIQUE_CARRIER in other files\n",
      "\tCARRIER_NAME: \t\tCarrier name, matches to UNIQUE_CARRIER_NAME or CARRIER_NAME in other files\n",
      "\tORIGIN_AIRPORT_ID: \tAirport ID, matches to ORIGIN_AIRPORT_ID in other files\n",
      "\tSERVICE_CLASS: \t\tService class of flight (required in download)\n",
      "\tREV_ACRFT_DEP_PERF_510: Departures performed for year\n",
      "\tREV_PAX_ENP_110: \tPassengers enplaned for year\n",
      "\n",
      "airports_list\n",
      "\tORIGIN_AIRPORT_ID: \tAirport ID, matches to ORIGIN_AIRPORT_ID in other files\n",
      "\tDISPLAY_AIRPORT_NAME: \tDisplay Airport, matches to DISPLAY_AIRPORT_NAME in other files\n",
      "\tORIGIN_CITY_NAME: \tCity\n",
      "\tNAME: \t\t\tMatches to NAME in airport_weather\n",
      "\n",
      "\n",
      "airport_weather_xxxx\n",
      "\tSee GHCND_documentation.pdf for full list\n",
      "\tImportant features:\n",
      "\tNAME: \t\t\tLocation of reading\n",
      "\tPRCP: \t\t\tInches of precipitation for day\n",
      "\tSNOW: \t\t\tInches of snowfall for day\n",
      "\tSNWD: \t\t\tInches of snow on ground for day\n",
      "\tTMAX: \t\t\tMax temperature for day\n",
      "\tAWND: \t\t\tMax wind speed for day\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Load the metadata (column descriptions)\n",
    "metadata = open(\"data/raw_data_documentation.txt\", \"r\")\n",
    "print(metadata.read())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>MONTH</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>DEP_DEL15</th>\n",
       "      <th>DEP_TIME_BLK</th>\n",
       "      <th>DISTANCE_GROUP</th>\n",
       "      <th>SEGMENT_NUMBER</th>\n",
       "      <th>CONCURRENT_FLIGHTS</th>\n",
       "      <th>NUMBER_OF_SEATS</th>\n",
       "      <th>CARRIER_NAME</th>\n",
       "      <th>AIRPORT_FLIGHTS_MONTH</th>\n",
       "      <th>...</th>\n",
       "      <th>PLANE_AGE</th>\n",
       "      <th>DEPARTING_AIRPORT</th>\n",
       "      <th>LATITUDE</th>\n",
       "      <th>LONGITUDE</th>\n",
       "      <th>PREVIOUS_AIRPORT</th>\n",
       "      <th>PRCP</th>\n",
       "      <th>SNOW</th>\n",
       "      <th>SNWD</th>\n",
       "      <th>TMAX</th>\n",
       "      <th>AWND</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>25</td>\n",
       "      <td>143</td>\n",
       "      <td>Southwest Airlines Co.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>8</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0700-0759</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>191</td>\n",
       "      <td>Delta Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>199</td>\n",
       "      <td>Delta Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>18</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>9</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>180</td>\n",
       "      <td>Delta Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0001-0559</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>182</td>\n",
       "      <td>Spirit Air Lines</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0001-0559</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>180</td>\n",
       "      <td>Frontier Airlines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0700-0759</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>186</td>\n",
       "      <td>Frontier Airlines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>0001-0559</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>186</td>\n",
       "      <td>Frontier Airlines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0001-0559</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>180</td>\n",
       "      <td>Frontier Airlines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>8</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>186</td>\n",
       "      <td>Frontier Airlines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>10 rows × 26 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   MONTH  DAY_OF_WEEK  DEP_DEL15 DEP_TIME_BLK  DISTANCE_GROUP  SEGMENT_NUMBER  \\\n",
       "0      1            7          0    0800-0859               2               1   \n",
       "1      1            7          0    0700-0759               7               1   \n",
       "2      1            7          0    0600-0659               7               1   \n",
       "3      1            7          0    0600-0659               9               1   \n",
       "4      1            7          0    0001-0559               7               1   \n",
       "5      1            7          0    0001-0559               3               1   \n",
       "6      1            7          0    0700-0759               6               1   \n",
       "7      1            7          1    0001-0559               7               1   \n",
       "8      1            7          0    0001-0559               7               1   \n",
       "9      1            7          0    0600-0659               8               1   \n",
       "\n",
       "   CONCURRENT_FLIGHTS  NUMBER_OF_SEATS            CARRIER_NAME  \\\n",
       "0                  25              143  Southwest Airlines Co.   \n",
       "1                  29              191    Delta Air Lines Inc.   \n",
       "2                  27              199    Delta Air Lines Inc.   \n",
       "3                  27              180    Delta Air Lines Inc.   \n",
       "4                  10              182        Spirit Air Lines   \n",
       "5                  10              180  Frontier Airlines Inc.   \n",
       "6                  29              186  Frontier Airlines Inc.   \n",
       "7                  10              186  Frontier Airlines Inc.   \n",
       "8                  10              180  Frontier Airlines Inc.   \n",
       "9                  27              186  Frontier Airlines Inc.   \n",
       "\n",
       "   AIRPORT_FLIGHTS_MONTH  ...  PLANE_AGE       DEPARTING_AIRPORT  LATITUDE  \\\n",
       "0                  13056  ...          8  McCarran International     36.08   \n",
       "1                  13056  ...          3  McCarran International     36.08   \n",
       "2                  13056  ...         18  McCarran International     36.08   \n",
       "3                  13056  ...          2  McCarran International     36.08   \n",
       "4                  13056  ...          1  McCarran International     36.08   \n",
       "5                  13056  ...          5  McCarran International     36.08   \n",
       "6                  13056  ...          2  McCarran International     36.08   \n",
       "7                  13056  ...          3  McCarran International     36.08   \n",
       "8                  13056  ...          3  McCarran International     36.08   \n",
       "9                  13056  ...          1  McCarran International     36.08   \n",
       "\n",
       "   LONGITUDE  PREVIOUS_AIRPORT  PRCP  SNOW SNWD  TMAX  AWND  \n",
       "0   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "1   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "2   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "3   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "4   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "5   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "6   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "7   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "8   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "9   -115.152              NONE   0.0   0.0  0.0  65.0  2.91  \n",
       "\n",
       "[10 rows x 26 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load the flight data\n",
    "data = pd.read_csv('data/full_data_flightdelay.csv')\n",
    "\n",
    "# Preview the first 10 records\n",
    "data.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataframe above gives me an initial look into the dataset, but I will apply a few more methods to gain a better understanding (i.e. `info()`, `isna()`, etc.). These techniques will help me learn more about my data, including the existence of any missing values and the data types of each column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 6489062 entries, 0 to 6489061\n",
      "Data columns (total 26 columns):\n",
      " #   Column                         Dtype  \n",
      "---  ------                         -----  \n",
      " 0   MONTH                          int64  \n",
      " 1   DAY_OF_WEEK                    int64  \n",
      " 2   DEP_DEL15                      int64  \n",
      " 3   DEP_TIME_BLK                   object \n",
      " 4   DISTANCE_GROUP                 int64  \n",
      " 5   SEGMENT_NUMBER                 int64  \n",
      " 6   CONCURRENT_FLIGHTS             int64  \n",
      " 7   NUMBER_OF_SEATS                int64  \n",
      " 8   CARRIER_NAME                   object \n",
      " 9   AIRPORT_FLIGHTS_MONTH          int64  \n",
      " 10  AIRLINE_FLIGHTS_MONTH          int64  \n",
      " 11  AIRLINE_AIRPORT_FLIGHTS_MONTH  int64  \n",
      " 12  AVG_MONTHLY_PASS_AIRPORT       int64  \n",
      " 13  AVG_MONTHLY_PASS_AIRLINE       int64  \n",
      " 14  FLT_ATTENDANTS_PER_PASS        float64\n",
      " 15  GROUND_SERV_PER_PASS           float64\n",
      " 16  PLANE_AGE                      int64  \n",
      " 17  DEPARTING_AIRPORT              object \n",
      " 18  LATITUDE                       float64\n",
      " 19  LONGITUDE                      float64\n",
      " 20  PREVIOUS_AIRPORT               object \n",
      " 21  PRCP                           float64\n",
      " 22  SNOW                           float64\n",
      " 23  SNWD                           float64\n",
      " 24  TMAX                           float64\n",
      " 25  AWND                           float64\n",
      "dtypes: float64(9), int64(13), object(4)\n",
      "memory usage: 1.3+ GB\n"
     ]
    }
   ],
   "source": [
    "# Print column information\n",
    "data.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I can see the method above displays the data type for each feature. Another point is the amount of data present. There are over 6.4 million observations, which will need to be reduced considering I'm only interested in United Airlines flights. The method below filters out the records based on the airline, revealing there are over 600,000 observations that will be analyzed to build my algorithm. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Southwest Airlines Co.          1296329\n",
       "Delta Air Lines Inc.             938346\n",
       "American Airlines Inc.           903640\n",
       "United Air Lines Inc.            601044\n",
       "SkyWest Airlines Inc.            584204\n",
       "Midwest Airline, Inc.            300154\n",
       "JetBlue Airways                  269596\n",
       "Alaska Airlines Inc.             239337\n",
       "American Eagle Airlines Inc.     228792\n",
       "Comair Inc.                      219324\n",
       "Endeavor Air Inc.                203827\n",
       "Spirit Air Lines                 189419\n",
       "Mesa Airlines Inc.               177600\n",
       "Frontier Airlines Inc.           120872\n",
       "Atlantic Southeast Airlines       99044\n",
       "Hawaiian Airlines Inc.            74898\n",
       "Allegiant Air                     42636\n",
       "Name: CARRIER_NAME, dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Print airline information\n",
    "data['CARRIER_NAME'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The first method also gives me an idea into whether there are any missing values in the dataset. To know for sure, I will apply some more methods that will take the sum of every missing value in each column and returns those values. I find that, fortunately, there are no missing values present in my data, which will make cleaning the data more straightforward."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MONTH                            0\n",
       "DAY_OF_WEEK                      0\n",
       "DEP_DEL15                        0\n",
       "DEP_TIME_BLK                     0\n",
       "DISTANCE_GROUP                   0\n",
       "SEGMENT_NUMBER                   0\n",
       "CONCURRENT_FLIGHTS               0\n",
       "NUMBER_OF_SEATS                  0\n",
       "CARRIER_NAME                     0\n",
       "AIRPORT_FLIGHTS_MONTH            0\n",
       "AIRLINE_FLIGHTS_MONTH            0\n",
       "AIRLINE_AIRPORT_FLIGHTS_MONTH    0\n",
       "AVG_MONTHLY_PASS_AIRPORT         0\n",
       "AVG_MONTHLY_PASS_AIRLINE         0\n",
       "FLT_ATTENDANTS_PER_PASS          0\n",
       "GROUND_SERV_PER_PASS             0\n",
       "PLANE_AGE                        0\n",
       "DEPARTING_AIRPORT                0\n",
       "LATITUDE                         0\n",
       "LONGITUDE                        0\n",
       "PREVIOUS_AIRPORT                 0\n",
       "PRCP                             0\n",
       "SNOW                             0\n",
       "SNWD                             0\n",
       "TMAX                             0\n",
       "AWND                             0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Find the amount of missing values in each column\n",
    "data.isna().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preparation\n",
    "Now that I've gotten an initial look, it's time to begin preparing the data for modeling. I will start by making a copy of the original dataset, then filtering the data to keep only flights taken with United Airlines. From my earlier observations, I can see that will greatly reduce the data from 6 million to a little over 600,000 records. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make a copy of the dataset\n",
    "data2 = data.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>MONTH</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>DEP_DEL15</th>\n",
       "      <th>DEP_TIME_BLK</th>\n",
       "      <th>DISTANCE_GROUP</th>\n",
       "      <th>SEGMENT_NUMBER</th>\n",
       "      <th>CONCURRENT_FLIGHTS</th>\n",
       "      <th>NUMBER_OF_SEATS</th>\n",
       "      <th>CARRIER_NAME</th>\n",
       "      <th>AIRPORT_FLIGHTS_MONTH</th>\n",
       "      <th>...</th>\n",
       "      <th>PLANE_AGE</th>\n",
       "      <th>DEPARTING_AIRPORT</th>\n",
       "      <th>LATITUDE</th>\n",
       "      <th>LONGITUDE</th>\n",
       "      <th>PREVIOUS_AIRPORT</th>\n",
       "      <th>PRCP</th>\n",
       "      <th>SNOW</th>\n",
       "      <th>SNWD</th>\n",
       "      <th>TMAX</th>\n",
       "      <th>AWND</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>25</td>\n",
       "      <td>173</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>25</td>\n",
       "      <td>142</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>22</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0900-0959</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>28</td>\n",
       "      <td>154</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>1000-1059</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>142</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>19</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>173</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0700-0759</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>154</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>21</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>154</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>20</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>9</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>173</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>11</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0001-0559</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>154</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>27</td>\n",
       "      <td>173</td>\n",
       "      <td>United Air Lines Inc.</td>\n",
       "      <td>13056</td>\n",
       "      <td>...</td>\n",
       "      <td>4</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>36.08</td>\n",
       "      <td>-115.152</td>\n",
       "      <td>NONE</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>10 rows × 26 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    MONTH  DAY_OF_WEEK  DEP_DEL15 DEP_TIME_BLK  DISTANCE_GROUP  \\\n",
       "21      1            7          0    0800-0859               2   \n",
       "22      1            7          0    0800-0859               3   \n",
       "23      1            7          0    0900-0959               7   \n",
       "24      1            7          1    1000-1059               3   \n",
       "25      1            7          0    0600-0659               7   \n",
       "26      1            7          0    0700-0759               1   \n",
       "27      1            7          0    0600-0659               2   \n",
       "28      1            7          0    0600-0659               9   \n",
       "29      1            7          0    0001-0559               5   \n",
       "30      1            7          0    0600-0659               3   \n",
       "\n",
       "    SEGMENT_NUMBER  CONCURRENT_FLIGHTS  NUMBER_OF_SEATS  \\\n",
       "21               1                  25              173   \n",
       "22               1                  25              142   \n",
       "23               1                  28              154   \n",
       "24               1                  29              142   \n",
       "25               1                  27              173   \n",
       "26               1                  29              154   \n",
       "27               1                  27              154   \n",
       "28               1                  27              173   \n",
       "29               1                  10              154   \n",
       "30               1                  27              173   \n",
       "\n",
       "             CARRIER_NAME  AIRPORT_FLIGHTS_MONTH  ...  PLANE_AGE  \\\n",
       "21  United Air Lines Inc.                  13056  ...          6   \n",
       "22  United Air Lines Inc.                  13056  ...         22   \n",
       "23  United Air Lines Inc.                  13056  ...          3   \n",
       "24  United Air Lines Inc.                  13056  ...         19   \n",
       "25  United Air Lines Inc.                  13056  ...          4   \n",
       "26  United Air Lines Inc.                  13056  ...         21   \n",
       "27  United Air Lines Inc.                  13056  ...         20   \n",
       "28  United Air Lines Inc.                  13056  ...         11   \n",
       "29  United Air Lines Inc.                  13056  ...          3   \n",
       "30  United Air Lines Inc.                  13056  ...          4   \n",
       "\n",
       "         DEPARTING_AIRPORT  LATITUDE  LONGITUDE  PREVIOUS_AIRPORT  PRCP  SNOW  \\\n",
       "21  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "22  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "23  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "24  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "25  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "26  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "27  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "28  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "29  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "30  McCarran International     36.08   -115.152              NONE   0.0   0.0   \n",
       "\n",
       "   SNWD  TMAX  AWND  \n",
       "21  0.0  65.0  2.91  \n",
       "22  0.0  65.0  2.91  \n",
       "23  0.0  65.0  2.91  \n",
       "24  0.0  65.0  2.91  \n",
       "25  0.0  65.0  2.91  \n",
       "26  0.0  65.0  2.91  \n",
       "27  0.0  65.0  2.91  \n",
       "28  0.0  65.0  2.91  \n",
       "29  0.0  65.0  2.91  \n",
       "30  0.0  65.0  2.91  \n",
       "\n",
       "[10 rows x 26 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Filter United Airlines's records and list the first 10\n",
    "data2 = data2.loc[data2['CARRIER_NAME'] == 'United Air Lines Inc.']\n",
    "data2.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next step is to remove all the columns I believe are unnecessary or irrelevant to my model. This will reduce my dimensions from 26 to 12, which includes my target column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 601044 entries, 21 to 6489030\n",
      "Data columns (total 12 columns):\n",
      " #   Column             Non-Null Count   Dtype  \n",
      "---  ------             --------------   -----  \n",
      " 0   MONTH              601044 non-null  int64  \n",
      " 1   DAY_OF_WEEK        601044 non-null  int64  \n",
      " 2   DEP_DEL15          601044 non-null  int64  \n",
      " 3   DEP_TIME_BLK       601044 non-null  object \n",
      " 4   DISTANCE_GROUP     601044 non-null  int64  \n",
      " 5   PLANE_AGE          601044 non-null  int64  \n",
      " 6   DEPARTING_AIRPORT  601044 non-null  object \n",
      " 7   PRCP               601044 non-null  float64\n",
      " 8   SNOW               601044 non-null  float64\n",
      " 9   SNWD               601044 non-null  float64\n",
      " 10  TMAX               601044 non-null  float64\n",
      " 11  AWND               601044 non-null  float64\n",
      "dtypes: float64(5), int64(5), object(2)\n",
      "memory usage: 59.6+ MB\n"
     ]
    }
   ],
   "source": [
    "# Drop unnecessary columns and display column information \n",
    "cols_to_drop = ['SEGMENT_NUMBER', 'NUMBER_OF_SEATS', 'FLT_ATTENDANTS_PER_PASS', 'AVG_MONTHLY_PASS_AIRPORT', \n",
    "                'GROUND_SERV_PER_PASS', 'AIRLINE_AIRPORT_FLIGHTS_MONTH', 'PREVIOUS_AIRPORT', 'AVG_MONTHLY_PASS_AIRLINE',\n",
    "                'LATITUDE', 'LONGITUDE', 'CARRIER_NAME', 'CONCURRENT_FLIGHTS', 'AIRPORT_FLIGHTS_MONTH', 'AIRLINE_FLIGHTS_MONTH']\n",
    "data2 = data2.drop(cols_to_drop, axis=1)\n",
    "data2.info()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that the columns have been reduced, I will look to do the same with the rows by dropping any duplicates that exist in my dataset. I will also rename some of the column names for easier comprehension, moving forward, and then shift the target column to the end of my dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "18105"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Find the number of duplicate records\n",
    "data2.duplicated().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Drop duplicates\n",
    "data2.drop_duplicates(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MONTH</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>DELAYED</th>\n",
       "      <th>DEP_TIME_BLK</th>\n",
       "      <th>DISTANCE_GROUP</th>\n",
       "      <th>PLANE_AGE</th>\n",
       "      <th>DEPARTING_AIRPORT</th>\n",
       "      <th>PRECIPITATION</th>\n",
       "      <th>SNOW</th>\n",
       "      <th>SNOW_ON_GROUND</th>\n",
       "      <th>MAX_TEMP_FOR_DAY</th>\n",
       "      <th>MAX_WIND_FOR_DAY</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>3</td>\n",
       "      <td>22</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0900-0959</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>1000-1059</td>\n",
       "      <td>3</td>\n",
       "      <td>19</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    MONTH  DAY_OF_WEEK  DELAYED DEP_TIME_BLK  DISTANCE_GROUP  PLANE_AGE  \\\n",
       "21      1            7        0    0800-0859               2          6   \n",
       "22      1            7        0    0800-0859               3         22   \n",
       "23      1            7        0    0900-0959               7          3   \n",
       "24      1            7        1    1000-1059               3         19   \n",
       "25      1            7        0    0600-0659               7          4   \n",
       "\n",
       "         DEPARTING_AIRPORT  PRECIPITATION  SNOW  SNOW_ON_GROUND  \\\n",
       "21  McCarran International            0.0   0.0             0.0   \n",
       "22  McCarran International            0.0   0.0             0.0   \n",
       "23  McCarran International            0.0   0.0             0.0   \n",
       "24  McCarran International            0.0   0.0             0.0   \n",
       "25  McCarran International            0.0   0.0             0.0   \n",
       "\n",
       "    MAX_TEMP_FOR_DAY  MAX_WIND_FOR_DAY  \n",
       "21              65.0              2.91  \n",
       "22              65.0              2.91  \n",
       "23              65.0              2.91  \n",
       "24              65.0              2.91  \n",
       "25              65.0              2.91  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Rename the columns for easier comprehension & list the first 5 records\n",
    "new_col_names = {'DEP_DEL15': 'DELAYED',\n",
    "                'PRCP': 'PRECIPITATION',\n",
    "                'SNWD': 'SNOW_ON_GROUND',\n",
    "                'TMAX': 'MAX_TEMP_FOR_DAY',\n",
    "                'AWND': 'MAX_WIND_FOR_DAY'}\n",
    "data2 = data2.rename(new_col_names, axis=1)\n",
    "data2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MONTH</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>DEP_TIME_BLK</th>\n",
       "      <th>DISTANCE_GROUP</th>\n",
       "      <th>PLANE_AGE</th>\n",
       "      <th>DEPARTING_AIRPORT</th>\n",
       "      <th>PRECIPITATION</th>\n",
       "      <th>SNOW</th>\n",
       "      <th>SNOW_ON_GROUND</th>\n",
       "      <th>MAX_TEMP_FOR_DAY</th>\n",
       "      <th>MAX_WIND_FOR_DAY</th>\n",
       "      <th>DELAYED</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0800-0859</td>\n",
       "      <td>3</td>\n",
       "      <td>22</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0900-0959</td>\n",
       "      <td>7</td>\n",
       "      <td>3</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>1000-1059</td>\n",
       "      <td>3</td>\n",
       "      <td>19</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>0600-0659</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>McCarran International</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>2.91</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    MONTH  DAY_OF_WEEK DEP_TIME_BLK  DISTANCE_GROUP  PLANE_AGE  \\\n",
       "21      1            7    0800-0859               2          6   \n",
       "22      1            7    0800-0859               3         22   \n",
       "23      1            7    0900-0959               7          3   \n",
       "24      1            7    1000-1059               3         19   \n",
       "25      1            7    0600-0659               7          4   \n",
       "\n",
       "         DEPARTING_AIRPORT  PRECIPITATION  SNOW  SNOW_ON_GROUND  \\\n",
       "21  McCarran International            0.0   0.0             0.0   \n",
       "22  McCarran International            0.0   0.0             0.0   \n",
       "23  McCarran International            0.0   0.0             0.0   \n",
       "24  McCarran International            0.0   0.0             0.0   \n",
       "25  McCarran International            0.0   0.0             0.0   \n",
       "\n",
       "    MAX_TEMP_FOR_DAY  MAX_WIND_FOR_DAY  DELAYED  \n",
       "21              65.0              2.91        0  \n",
       "22              65.0              2.91        0  \n",
       "23              65.0              2.91        0  \n",
       "24              65.0              2.91        1  \n",
       "25              65.0              2.91        0  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shift the target column to the end \n",
    "cols_at_end = ['DELAYED']\n",
    "data2 = data2[[col for col in data2 if col not in cols_at_end] \n",
    "        + [col for col in cols_at_end if col in data2]]\n",
    "\n",
    "# Preview the first 5 records to confirm the change\n",
    "data2.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that the dataset has been cleaned up, I want to take a look at the distributions of both the features and the target. This will give me an idea of whether a class imbalance exists (which I suspect there does), and how the feature data is distributed within that imbalance. I'll start by getting a count of the records, based on the `'MONTH'` and `'DAY_OF_WEEK'` features. I plot these value counts, but I divide them based on the target variable - labeling the bars as `'Not Delayed'` or `'Delayed'`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1     43189\n",
       "2     40457\n",
       "3     49824\n",
       "4     49038\n",
       "5     51038\n",
       "6     50529\n",
       "7     52122\n",
       "8     52538\n",
       "9     47786\n",
       "10    51477\n",
       "11    46880\n",
       "12    48061\n",
       "Name: MONTH, dtype: int64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Break down the records by month\n",
    "data2['MONTH'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    86421\n",
       "2    85043\n",
       "3    84471\n",
       "4    84973\n",
       "5    86153\n",
       "6    73156\n",
       "7    82722\n",
       "Name: DAY_OF_WEEK, dtype: int64"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Break down the records by days of the week\n",
    "data2['DAY_OF_WEEK'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot feature distributions\n",
    "\n",
    "# Create month and days list objects\n",
    "months = ['January', 'February', 'March', 'April', 'May', 'June',\n",
    "         'July', 'August', 'September', 'October', 'November', 'December']\n",
    "days = ['Sunday', 'Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday']\n",
    "\n",
    "# Visualize the delay status by month\n",
    "fig, ax = plt.subplots(1, 2, figsize=(18, 12))\n",
    "plot1 = sns.histplot(data2, x='MONTH', hue='DELAYED', ax=ax[0], palette='bright', discrete=True)\n",
    "\n",
    "# Change the legend labels\n",
    "new_title = 'Delay Status'\n",
    "plot1.legend_.set_title(new_title)\n",
    "new_labels = ['Not Delayed', 'Delayed']\n",
    "for t, l in zip(plot1.legend_.texts, new_labels):\n",
    "    t.set_text(l)\n",
    "    \n",
    "# Add minor gridlines\n",
    "minor_locator = AutoMinorLocator(5)\n",
    "ax[0].yaxis.set_minor_locator(minor_locator)\n",
    "ax[0].set_axisbelow(True)\n",
    "plt.grid(which='both')\n",
    "ax[0].tick_params(which=\"both\", bottom=True)\n",
    "\n",
    "# Change x-tick labels to months (written form) and rotate the labels\n",
    "old_labels = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]\n",
    "ax[0].set_xticks(old_labels)\n",
    "ax[0].set_xticklabels(months)\n",
    "ax[0].tick_params(axis='x', labelrotation=35)\n",
    "ax[0].set_xlabel(\"Months\")\n",
    "ax[0].set_title(\"The Number of Delayed and Not-Delayed Flights by Month\")\n",
    "    \n",
    "# Visualize the delay status by days of the week    \n",
    "plot2 = sns.histplot(data2, x='DAY_OF_WEEK', hue='DELAYED', ax=ax[1], palette='bright', discrete=True)\n",
    "\n",
    "# Change the legend labels\n",
    "new_title = 'Delay Status'\n",
    "plot2.legend_.set_title(new_title)\n",
    "new_labels = ['Not Delayed', 'Delayed']\n",
    "for t, l in zip(plot2.legend_.texts, new_labels):\n",
    "    t.set_text(l)\n",
    "\n",
    "# Add minor gridlines\n",
    "minor_locator = AutoMinorLocator(5)\n",
    "ax[1].yaxis.set_minor_locator(minor_locator)\n",
    "ax[1].set_axisbelow(True)\n",
    "plt.grid(which='both')\n",
    "\n",
    "# Change x-tick labels to days (written form)\n",
    "old_labels = [1, 2, 3, 4, 5, 6, 7]\n",
    "ax[1].set_xticks(old_labels)\n",
    "ax[1].set_xticklabels(days)\n",
    "ax[1].set_xlabel(\"Days of the Week\")\n",
    "ax[1].set_title(\"The Number of Delayed and Not-Delayed Flights by Day of the Week\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on the visuals above, I see that there definitely exists a class imbalance in the dataset. In regard to the features, the first plot suggests that the summer months tend to see the highest number of delays, while the colder months seem to see less. In the second plot, the delays are highest on Wednesdays; although, the data seems to have low variation between the days of the week. \n",
    "\n",
    "The next few features are the `'DEP_TIME_BLK'` and the `'DISTANCE_GROUP'` columns. The first feature lists the time blocks for United Airlines flights, ranging from midnight (`'0001'`) to 11:59pm, and separated into hour blocks (except for the first block). The second feature lists all 11 distance groups, where the first group travels the shortest distance, and the eleventh group travels the farthest. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0001-0559    10189\n",
       "0600-0659    42020\n",
       "0700-0759    47223\n",
       "0800-0859    42906\n",
       "0900-0959    36314\n",
       "1000-1059    35549\n",
       "1100-1159    34927\n",
       "1200-1259    35124\n",
       "1300-1359    28503\n",
       "1400-1459    33245\n",
       "1500-1559    30224\n",
       "1600-1659    35246\n",
       "1700-1759    35880\n",
       "1800-1859    36416\n",
       "1900-1959    33797\n",
       "2000-2059    21716\n",
       "2100-2159    19542\n",
       "2200-2259    12919\n",
       "2300-2359    11199\n",
       "Name: DEP_TIME_BLK, dtype: int64"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Break down the records by departure time blocks\n",
    "data2['DEP_TIME_BLK'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1      29818\n",
       "2      66963\n",
       "3      80728\n",
       "4     112096\n",
       "5      65293\n",
       "6      46784\n",
       "7      60650\n",
       "8      23843\n",
       "9      16293\n",
       "10     50679\n",
       "11     29792\n",
       "Name: DISTANCE_GROUP, dtype: int64"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Break down the records by distance group\n",
    "data2['DISTANCE_GROUP'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1296x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot feature distributions\n",
    "fig, ax = plt.subplots(1, 2, figsize=(18, 12))\n",
    "\n",
    "# Visualize the delay status by departure time block\n",
    "plot3 = sns.histplot(data2, x='DEP_TIME_BLK', hue='DELAYED', ax=ax[0], palette='bright', discrete=True)\n",
    "\n",
    "# Change legend labels\n",
    "new_title = 'Delay Status'\n",
    "plot3.legend_.set_title(new_title)\n",
    "new_labels = ['Not Delayed', 'Delayed']\n",
    "for t, l in zip(plot3.legend_.texts, new_labels):\n",
    "    t.set_text(l)\n",
    "\n",
    "# Add minor gridlines\n",
    "minor_locator = AutoMinorLocator(5)\n",
    "ax[0].yaxis.set_minor_locator(minor_locator)\n",
    "ax[0].set_axisbelow(True)\n",
    "plt.grid(which='both')\n",
    "\n",
    "# Rotate x-tick labels\n",
    "ax[0].tick_params(axis='x', labelrotation=75)\n",
    "ax[0].tick_params(which=\"both\", bottom=True)\n",
    "ax[0].set_xlabel(\"Departure Time Blocks\")\n",
    "ax[0].set_title(\"The Number of Delayed and Not-Delayed Flights by Departure Block\")\n",
    "    \n",
    "# Visualize the delay status by distance group  \n",
    "plot4 = sns.histplot(data2, x='DISTANCE_GROUP', hue='DELAYED', ax=ax[1], palette='bright', discrete=True)\n",
    "\n",
    "# Change legend labels\n",
    "new_title = 'Delay Status'\n",
    "plot4.legend_.set_title(new_title)\n",
    "new_labels = ['Not Delayed', 'Delayed']\n",
    "for t, l in zip(plot4.legend_.texts, new_labels):\n",
    "    t.set_text(l)\n",
    "    \n",
    "# Add minor gridlines\n",
    "minor_locator = AutoMinorLocator(5)\n",
    "ax[1].yaxis.set_minor_locator(minor_locator)\n",
    "ax[1].set_axisbelow(True)\n",
    "plt.grid(which='both')\n",
    "\n",
    "# Change x-tick labels to distance group (written form)\n",
    "num_xlabels = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]\n",
    "str_xlabels = ['First', 'Second', 'Third', 'Fourth', 'Fifth', 'Sixth', 'Seventh', 'Eighth', 'Ninth', 'Tenth', 'Eleventh']\n",
    "ax[1].set_xticks(num_xlabels)\n",
    "ax[1].set_xticklabels(str_xlabels)\n",
    "ax[1].set_xlabel(\"Distance Group\")\n",
    "ax[1].set_title(\"The Number of Delayed and Not-Delayed Flights by Distance Group\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From what I see, the later time blocks (i.e., starting at 4pm to about 8pm) face the most delays. However, in the second visual, the third and fourth distance groups experience more delays, which is far less than those in higher distance groups. \n",
    "\n",
    "To be able to plot the rest of the features, I use both kdeplots and boxplots because the data within those columns are numerical and continuous in nature. I see that all six graphs contain outliers that affect the distribution one way or another, as well. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x864 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the remaining continuous variables with the delay status \n",
    "fig, ax = plt.subplots(3, 2, figsize=(16, 12))\n",
    "kde1 = sns.kdeplot(data=data2, x=\"PRECIPITATION\", hue=\"DELAYED\", ax=ax[0,0])\n",
    "kde2 = sns.kdeplot(data=data2, x=\"SNOW\", hue=\"DELAYED\", ax=ax[0,1])\n",
    "kde3 = sns.kdeplot(data=data2, x=\"SNOW_ON_GROUND\", hue=\"DELAYED\", ax=ax[1,0])\n",
    "box1 = sns.boxplot(data=data2, x=\"PLANE_AGE\", y=\"DELAYED\", orient='h', ax=ax[1,1])\n",
    "box2 = sns.boxplot(data=data2, x=\"MAX_TEMP_FOR_DAY\", y=\"DELAYED\", orient='h', ax=ax[2,0])\n",
    "box3 = sns.boxplot(data=data2, x=\"MAX_WIND_FOR_DAY\", y=\"DELAYED\", orient='h', ax=ax[2,1]);\n",
    "\n",
    "# Change the legend in the kdeplots for better comprehension\n",
    "kdeplots = [kde1, kde2, kde3]\n",
    "for plot in kdeplots:\n",
    "    new_title = 'Delay Status'\n",
    "    plot.legend_.set_title(new_title)\n",
    "    new_labels = ['Not Delayed', 'Delayed']\n",
    "    for t, l in zip(plot.legend_.texts, new_labels):\n",
    "        t.set_text(l)\n",
    "    plot.set_xlim(-1, 6) \n",
    "\n",
    "# Change the y-axis label for the boxplots\n",
    "boxplots = [box1, box2, box3]\n",
    "for boxplot in boxplots:\n",
    "    boxplot.set(ylabel='Delay Status')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It's time to examine the target, and confirm the class imbalance I mentioned earlier. First, I will plot my target as a bar plot, then as a pie chart. I plot both to show both the count and weight of each label in the dataset, respectively. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the class (target) distribution\n",
    "fig, ax = plt.subplots(figsize=(12,10))\n",
    "\n",
    "# Add minor gridlines\n",
    "minor_locator = AutoMinorLocator(5)\n",
    "ax.yaxis.set_minor_locator(minor_locator)\n",
    "ax.set_axisbelow(True)\n",
    "plt.grid(which='both')\n",
    "\n",
    "# Plot the target \n",
    "sns.countplot(data=data2, x='DELAYED', orient='v')\n",
    "ax.set_title('Target Distribution')\n",
    "\n",
    "# Change x-tick labels and axis labels\n",
    "old_labels = [0, 1]\n",
    "new_xlabels = ['Not Delayed', 'Delayed']\n",
    "ax.set_xticks(old_labels)\n",
    "ax.set_xticklabels(new_xlabels)\n",
    "ax.set_xlabel(\"Flight Status\")\n",
    "ax.set_ylabel(\"Count\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The bar plot above further confirms my speculation of a class imbalance - showing over 460,000 flights as `'Not Delayed'` and a little under 120,000 as `'Delayed'` flights. The pie plot below reinforces this observation, but displays the weight of each label in the data. I see that out of 600,000 flights, close to 20% are delayed - this is quite high considering the high amount of flights taken in a year through United. According to the Bureau of Transportation, 18.72% of flights in 2019 were delayed (all carriers included), meaning United exceeded the industry average."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a pie plot to further visualize the class distribution\n",
    "fig, ax = plt.subplots(figsize=(12, 10))\n",
    "data2['DELAYED'].value_counts().plot.pie(autopct='%.2f', title='Class Distribution', \n",
    "                                        labels=['Not Delayed', 'Delayed'], colors=['orangered', 'royalblue'])\n",
    "ax.yaxis.set_visible(False);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I feel I've gained sufficient insight into how my data is structured and distributed; so, now it's time to begin modeling and building my algorithm. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Further Preprocessing:\n",
    "In order to properly model my dataframe, I need to encode the categorical features into quanitative data. I can call scikit-learn's `OneHotEncoder` function to encode my five categorical columns into multiple, numerical columns; then list the shapes of all datasets to confirm the transformation. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset above will serve as my final dataframe, but I will also divide it into a training, testing and validation set. I will split the data 75/25, with 25% of the dataset reserved for the test set, and a seed set for reproducibility. Now, I will split the test data even further (by half, actually) into a smaller test set and a newly-formed validation set to iterate through my modeling process. I check the shape multiple times to be sure the data stays intact. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(437204, 11)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Split the dataset into training and testing sets\n",
    "y = data2['DELAYED']\n",
    "X = data2.drop('DELAYED', axis=1)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.25, random_state=seed)\n",
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(72868, 11)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Split the test dataset in half to create a validation dataset\n",
    "X_test, X_val, y_test, y_val = train_test_split(X_test, y_test, test_size=.5, random_state=seed)\n",
    "X_val.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that the data's been split, I can begin encoding each dataset separately. The test set will act as my \"unseen data\" and will be reserved for my final evaluation. As stated earlier, I will list the shapes just for peace of mind. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Encode categorical features\n",
    "cols_to_encode = ['MONTH', 'DAY_OF_WEEK', 'DEP_TIME_BLK', 'DISTANCE_GROUP', 'DEPARTING_AIRPORT']\n",
    "ohe = OneHotEncoder(handle_unknown='ignore')\n",
    "ohe.fit(X_train[cols_to_encode])\n",
    "new_train_df = pd.DataFrame(ohe.transform(X_train[cols_to_encode]).todense(), \n",
    "                            columns=ohe.get_feature_names(), index=X_train.index)\n",
    "X_train = pd.concat([X_train, new_train_df], axis=1).drop(cols_to_encode, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_df = pd.DataFrame(ohe.transform(X_val[cols_to_encode]).todense(), \n",
    "                      columns=ohe.get_feature_names(), index=X_val.index)\n",
    "X_val = pd.concat([X_val.drop(cols_to_encode, axis=1), val_df], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df = pd.DataFrame(ohe.transform(X_test[cols_to_encode]).todense(), \n",
    "                      columns=ohe.get_feature_names(), index=X_test.index)\n",
    "X_test = pd.concat([X_test.drop(cols_to_encode, axis=1), test_df], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(437204, 138)\n",
      "(72868, 138)\n",
      "(72867, 138)\n"
     ]
    }
   ],
   "source": [
    "# List the first five records to confirm the transformation\n",
    "print(X_train.shape)\n",
    "print(X_val.shape)\n",
    "print(X_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Given the initial class imbalance, I need to apply a sampling technique - specifically, oversampling (SMOTE) - to augment the minority class (`'Delayed'`) in my target to an even ratio with the majority class (`'Not Delayed'`). I use the `Counter` function to show this change. I also plot the pie chart to display the new class weights, but this time expecting an even ratio."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Before sampling:  Counter({0: 350513, 1: 86691})\n",
      "After sampling:  Counter({0: 350513, 1: 350513})\n"
     ]
    }
   ],
   "source": [
    "# Count the number of classes before oversampling\n",
    "counter = Counter(y_train)\n",
    "print('Before sampling: ', counter)\n",
    "\n",
    "# Instantiate the SMOTE function and oversample the training data\n",
    "smt = SMOTE(random_state=seed) \n",
    "X_train_res, y_train_res = smt.fit_resample(X_train, y_train)\n",
    "\n",
    "# Count the number of classes after oversampling\n",
    "counter_res = Counter(y_train_res)\n",
    "print('After sampling: ', counter_res)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the balanced class distribution\n",
    "fig, ax = plt.subplots(figsize=(12, 10))\n",
    "y_train_res.value_counts().plot.pie(autopct='%.2f', title='Class Distribution in the Training Data', \n",
    "                                        labels=['Not Delayed', 'Delayed'], colors=['orangered', 'royalblue'])\n",
    "ax.yaxis.set_visible(False);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As shown above, the SMOTE method worked! To see just how much my training data has grown, I print the shape of the training data below. I see that the dataset now has **701,026 observations** to train my model on, and that the features remained intact. This increase in observations should improve my modeling performance and help me avoid any issues or controversy in results. The downside to this is the long run times I'm sure to experience due to my computational constraints."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(701026, 138)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Print the balanced training data's shape\n",
    "X_train_res.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling\n",
    "\n",
    "Now it's time to begin modeling! I will create a few baseline models, to start, then cross-validate 4 different classifiers, and evaluate them based on precision, recall, and f1 scores. As a reminder, my main focus is to improve precision, which effects the number of false positive predictions. The false positive count, in this case, represents the amount of on-time flights predicted as delays. This could be detrimental if not addressed properly; the model could spread misinformation, and cause passengers to miss their on-time flights. This would lead to further client disapproval, and hurt United's sales and reputation. However, I will still track the recall score, and the f1 score, which is the harmonic mean of precision and recall. The baseline model with the best trio of average metric scores will be selected to undergo hyperparameter tuning and further evaluation.\n",
    "\n",
    "### Baseline Modeling:\n",
    "\n",
    "I will begin by using scikit-learn's `make_pipeline` function in order to pass a scaler and a classifier into my pipeline(s). I use this function over the library's long form version to avoid naming the estimators. I cross-validate these pipelines in 3 folds, then take the average of each score and print them for each classifier. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation precision, recall and f1 scores for Decision Tree: [0.8416852479470313, 0.7756237594173561, 0.7595007602911071]\n",
      "\n",
      "Validation precision, recall and f1 scores for Random Forest: [0.9313695562886423, 0.7666883474008875, 0.7864523707153354]\n",
      "\n",
      "Validation precision, recall and f1 scores for AdaBoost: [0.8480503638132815, 0.691946367820922, 0.715686314340466]\n",
      "\n",
      "Validation precision, recall and f1 scores for Gradient Boosting: [0.9371740386968425, 0.7141624780031419, 0.7516701938488651]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Make two lists, one containing classifier names and the other containing the classifiers\n",
    "classifier_names = ['Decision Tree', 'Random Forest', 'AdaBoost', 'Gradient Boosting']\n",
    "classifiers = [DecisionTreeClassifier(random_state=seed), RandomForestClassifier(random_state=seed), \n",
    "               AdaBoostClassifier(random_state=seed), GradientBoostingClassifier(random_state=seed)]\n",
    "\n",
    "# Loop through the two lists to create a pipeline, cross-validate the model, and print the relevant scores\n",
    "for name, classifier in zip(classifier_names, classifiers):\n",
    "    pipe = make_pipeline((MinMaxScaler()),\n",
    "                          (classifier))\n",
    "    score = cross_validate(pipe, X_train_res, y_train_res, cv=3, scoring=['precision', 'recall', 'f1'])\n",
    "    model_scores = []\n",
    "    keys = ['test_precision', 'test_recall', 'test_f1']\n",
    "    for key in keys:\n",
    "        model_scores.append(np.mean(score[key]))\n",
    "    print(\"Validation precision, recall and f1 scores for {}: {}\\n\".format(name, model_scores))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on the scores listed above, it seems that the Random Forest classifier outperforms the field, which eliminates the other models from further iteration. The Random Forest classifier is 93.1% precise, which is slightly lower than the Gradient Boosting model's score, but negligible enough that we can still go with the former. We can also justify this decision by taking a look at the two other metrics - the f1 score & the recall. As shown above, the Random Forest's set of scores are the highest among the four; thus, we can feel safe about selecting it as the \"best\" model for further analysis.  \n",
    "\n",
    "93.1% validation precision is fantastic as a baseline, but also could be a sign that my model is overfitting. My next step will be trying to improve this start with some hyperparameter tuning. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tuning the \"Best\" Baseline Model:\n",
    "\n",
    "In order to keep in line with the business objectives, I need to set the `scoring` parameter to the three metrics I have been tracking so far. I will create a dictionary object that holds these three scores, and pass it into the `GridSearchCV` function. I will also pass a pipeline (long form this time) into the function, that will apply `MinMaxScaler` and my chosen classifier over the folds (in this case, 3) from the exhaustive grid search process. Because I'm using scikit-learn's long form pipeline function, it's vital I name each estimator in the steps parameter. My last step before fitting the grid search object is to create a parameter grid for the function to circulate through. It's important I don't forget to put the classifier name I choose as a prefix followed by two underscores before each parameter name in order to avoid error. Once I've set all the relevant parameters, I fit the grid search object.\n",
    "\n",
    ">Note: Due to the number of fits, and depending on the computational constraints, this code can take over **3 hours** to fully run!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a dictionary of relevant scoring metrics\n",
    "scorers = {'precision_score': make_scorer(precision_score), \n",
    "           'recall_score': make_scorer(recall_score),\n",
    "           'f1_score': make_scorer(f1_score)}\n",
    "\n",
    "# Create a Pipeline with a scaler and classifier step\n",
    "rf_pipe = Pipeline(steps=[('scaler', MinMaxScaler()),\n",
    "                          ('classifier', RandomForestClassifier(random_state=seed))])\n",
    "\n",
    "# Create a parameter grid for the Random Forest Classifier\n",
    "rf_param_grid = {'classifier__n_estimators': [50, 100, 125], \n",
    "                 'classifier__max_depth': [None, 10, 20], \n",
    "                 'classifier__min_samples_leaf': [5, 10, 15]}\n",
    "\n",
    "# Instantiate the GridSearchCV function with the 'refit' parameter set to 'precision_score'\n",
    "rf_gridsearch = GridSearchCV(rf_pipe, rf_param_grid, scoring=scorers,\n",
    "                              refit='precision_score', verbose=3, cv=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 27 candidates, totalling 81 fits\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.423, precision_score=0.989, recall_score=0.269, total= 1.5min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:  1.5min remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.961, precision_score=0.941, recall_score=0.982, total= 1.7min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:  3.2min remaining:    0.0s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.960, precision_score=0.941, recall_score=0.981, total= 1.7min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.422, precision_score=0.990, recall_score=0.268, total= 3.0min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.962, precision_score=0.942, recall_score=0.982, total= 3.3min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.962, precision_score=0.943, recall_score=0.982, total= 3.3min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.423, precision_score=0.990, recall_score=0.269, total= 3.8min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.962, precision_score=0.942, recall_score=0.983, total= 4.1min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.962, precision_score=0.942, recall_score=0.982, total= 4.1min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.418, precision_score=0.988, recall_score=0.265, total= 1.5min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.953, precision_score=0.935, recall_score=0.971, total= 1.6min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.953, precision_score=0.937, recall_score=0.971, total= 1.6min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.416, precision_score=0.989, recall_score=0.263, total= 2.9min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.955, precision_score=0.938, recall_score=0.973, total= 3.1min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.955, precision_score=0.938, recall_score=0.972, total= 3.1min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.416, precision_score=0.989, recall_score=0.264, total= 3.7min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.955, precision_score=0.937, recall_score=0.973, total= 3.8min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.955, precision_score=0.939, recall_score=0.973, total= 3.8min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.417, precision_score=0.985, recall_score=0.265, total= 1.4min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.949, precision_score=0.932, recall_score=0.966, total= 1.5min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.948, precision_score=0.932, recall_score=0.965, total=299.1min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.416, precision_score=0.987, recall_score=0.263, total= 2.7min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.949, precision_score=0.933, recall_score=0.966, total=17.5min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.950, precision_score=0.934, recall_score=0.966, total= 2.8min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.416, precision_score=0.986, recall_score=0.264, total= 3.4min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.950, precision_score=0.933, recall_score=0.967, total= 3.6min\n",
      "[CV] classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=None, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.950, precision_score=0.934, recall_score=0.967, total= 3.6min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.541, precision_score=0.802, recall_score=0.409, total=  37.4s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.870, precision_score=0.818, recall_score=0.931, total=  36.0s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.873, precision_score=0.822, recall_score=0.929, total=  36.8s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.533, precision_score=0.815, recall_score=0.396, total=  40.5s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.871, precision_score=0.817, recall_score=0.933, total=  55.5s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.874, precision_score=0.822, recall_score=0.934, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.535, precision_score=0.811, recall_score=0.400, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.871, precision_score=0.821, recall_score=0.929, total= 1.4min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.873, precision_score=0.823, recall_score=0.930, total= 1.4min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.545, precision_score=0.792, recall_score=0.416, total=  37.4s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.866, precision_score=0.810, recall_score=0.930, total=  36.1s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.865, precision_score=0.809, recall_score=0.930, total=  36.2s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.538, precision_score=0.807, recall_score=0.404, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.870, precision_score=0.816, recall_score=0.932, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.869, precision_score=0.814, recall_score=0.932, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.537, precision_score=0.807, recall_score=0.403, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.871, precision_score=0.819, recall_score=0.928, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.870, precision_score=0.818, recall_score=0.929, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.537, precision_score=0.800, recall_score=0.404, total=  39.5s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.869, precision_score=0.820, recall_score=0.924, total=  37.1s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.875, precision_score=0.829, recall_score=0.926, total=  36.3s\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.535, precision_score=0.809, recall_score=0.400, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.870, precision_score=0.817, recall_score=0.930, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.876, precision_score=0.828, recall_score=0.929, total= 1.2min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.533, precision_score=0.811, recall_score=0.397, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.871, precision_score=0.821, recall_score=0.928, total= 1.5min\n",
      "[CV] classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=10, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.875, precision_score=0.829, recall_score=0.926, total= 1.5min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.471, precision_score=0.923, recall_score=0.316, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.929, precision_score=0.905, recall_score=0.954, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=50, f1_score=0.926, precision_score=0.901, recall_score=0.954, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.464, precision_score=0.938, recall_score=0.309, total= 2.2min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.932, precision_score=0.909, recall_score=0.956, total= 2.2min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=100, f1_score=0.931, precision_score=0.908, recall_score=0.957, total= 2.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.465, precision_score=0.938, recall_score=0.309, total= 2.7min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.932, precision_score=0.909, recall_score=0.956, total= 2.8min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=5, classifier__n_estimators=125, f1_score=0.932, precision_score=0.908, recall_score=0.958, total= 2.7min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.469, precision_score=0.921, recall_score=0.314, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.925, precision_score=0.902, recall_score=0.949, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=50, f1_score=0.925, precision_score=0.901, recall_score=0.950, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.460, precision_score=0.936, recall_score=0.305, total= 2.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.927, precision_score=0.904, recall_score=0.952, total= 2.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=100, f1_score=0.927, precision_score=0.903, recall_score=0.953, total= 2.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.462, precision_score=0.935, recall_score=0.307, total= 2.7min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.927, precision_score=0.903, recall_score=0.952, total= 2.7min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=10, classifier__n_estimators=125, f1_score=0.928, precision_score=0.904, recall_score=0.954, total= 2.7min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.472, precision_score=0.916, recall_score=0.318, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.922, precision_score=0.898, recall_score=0.948, total= 1.0min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=50, f1_score=0.922, precision_score=0.898, recall_score=0.947, total= 1.1min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.464, precision_score=0.931, recall_score=0.309, total= 2.5min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.925, precision_score=0.900, recall_score=0.951, total= 2.3min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=100, f1_score=0.925, precision_score=0.901, recall_score=0.950, total= 2.3min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.464, precision_score=0.931, recall_score=0.309, total= 2.9min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.924, precision_score=0.900, recall_score=0.950, total= 2.9min\n",
      "[CV] classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125 \n",
      "[CV]  classifier__max_depth=20, classifier__min_samples_leaf=15, classifier__n_estimators=125, f1_score=0.925, precision_score=0.901, recall_score=0.951, total= 3.7min\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done  81 out of  81 | elapsed: 472.1min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3,\n",
       "             estimator=Pipeline(steps=[('scaler', MinMaxScaler()),\n",
       "                                       ('classifier',\n",
       "                                        RandomForestClassifier(random_state=24))]),\n",
       "             param_grid={'classifier__max_depth': [None, 10, 20],\n",
       "                         'classifier__min_samples_leaf': [5, 10, 15],\n",
       "                         'classifier__n_estimators': [50, 100, 125]},\n",
       "             refit='precision_score',\n",
       "             scoring={'f1_score': make_scorer(f1_score),\n",
       "                      'precision_score': make_scorer(precision_score),\n",
       "                      'recall_score': make_scorer(recall_score)},\n",
       "             verbose=3)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fit the GridSearchCV object\n",
    "rf_gridsearch.fit(X_train_res, y_train_res)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that the grid search has fully run, I can use the `best_params_` and `best_score_` attributes to find just that - the best parameters found through my grid search and the best cross-validated score of the best estimator, respectively. I see that my grid search determined 100 estimators and a minimum number of 5 samples per leaf in the Random Forest classifier were best, and refit the model based on these hyperparameters. The best score (0.931) is pretty high, and could be overfitting. I plot the confusion matrix to get a better look into how well this model does."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params for Random Forest Classifier refit for precision_score:\n",
      "{'classifier__max_depth': None, 'classifier__min_samples_leaf': 5, 'classifier__n_estimators': 100}\n",
      "\n",
      "\n",
      "Best score for Random Forest Classifier refit for precision_score:\n",
      "0.9583924767042996\n"
     ]
    }
   ],
   "source": [
    "# Print the 'best' parameters and the best score for the model\n",
    "print('Best params for Random Forest Classifier refit for {}:'.format('precision_score'))\n",
    "print(rf_gridsearch.best_params_)\n",
    "print('\\n')\n",
    "print('Best score for Random Forest Classifier refit for {}:'.format('precision_score'))\n",
    "print(rf_gridsearch.best_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the confusion matrix\n",
    "y_hat_val = rf_gridsearch.predict(X_val)\n",
    "plt.figure(figsize=(10, 8))\n",
    "sns.heatmap(confusion_matrix(y_val, y_hat_val), annot=True,\n",
    "           fmt='.1f', xticklabels=['Not Delayed', 'Delayed'],\n",
    "           yticklabels=['Not Delayed', 'Delayed'], cmap='OrRd', cbar=False)\n",
    "plt.xlabel(\"Predicted Labels\")\n",
    "plt.ylabel(\"True Labels\")\n",
    "plt.title('Confusion Matrix of the Optimized Random Forest On the Validation Data')\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From initial observation, I can see this model needs more training, especially if my hope is to reduce the false positive count while also predicting delays. The false positive count is 1,505, which is a decent start; however, there are 12,333 predictions misclassified as not delayed, which is a problem. I print the classification report below to get a full look into the grid search's metric scores, including recall and f1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.82      0.97      0.89     58461\n",
      "           1       0.58      0.14      0.23     14407\n",
      "\n",
      "    accuracy                           0.81     72868\n",
      "   macro avg       0.70      0.56      0.56     72868\n",
      "weighted avg       0.77      0.81      0.76     72868\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the classification report\n",
    "print(classification_report(y_val, y_hat_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The scores above (58.0%, 14.0%, 23.0%, respectively) is pretty low and invite more work. Due to constraints, I will accept these scores as the machine learning standard for comparison against another technique: deep learning. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Deep Learning:\n",
    "\n",
    "Before I begin creating neural networks to train my data on, it's important I normalize the input datasets so that they are all on the same scale; otherwise, the models will generate flawed results. I will use `MinMaxScaler` once again, and transform all three input datasets, but only fitting on the training data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(701026, 138)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Use the MinMaxScaler function to scale the data\n",
    "cols = X_train_res.columns\n",
    "\n",
    "mmscaler = MinMaxScaler()\n",
    "X_train_scaled = mmscaler.fit_transform(X_train_res)\n",
    "X_val_scaled = mmscaler.transform(X_val)\n",
    "X_test_scaled = mmscaler.transform(X_test)\n",
    "\n",
    "X_train = pd.DataFrame(X_train_scaled, columns=cols)\n",
    "X_val = pd.DataFrame(X_val_scaled, columns=cols)\n",
    "X_test = pd.DataFrame(X_test_scaled, columns=cols)\n",
    "\n",
    "# Check & reconfirm the training data's shape\n",
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I see from the last line of code that the shape of the training data has remained intact, and that my data is ready to be modeled. To begin, I build a small baseline neural network - specifically, with 2 hidden layers and 1 output layer). I also set some regularization to help counter any potential overfitting to start. \n",
    "\n",
    "Seeing as how this is a classification problem, it's important I select loss and optimizer functions that work well in binary classification. From my knowledge, binary cross-entropy and Adam optimization work best in these situations, so I pass them along with a metrics parameter in the compiler function. The summary below shows the number of parameters that exist (and will be trained) in the network, which in this case is a little under 10,000 parameters!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 64)                8896      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 16)                1040      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 17        \n",
      "=================================================================\n",
      "Total params: 9,953\n",
      "Trainable params: 9,953\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Create a small-layered baseline neural network\n",
    "neural_network = Sequential()\n",
    "neural_network.add(Dense(64, activation='relu', \n",
    "                            kernel_regularizer=regularizers.l2(0.01), input_shape=(138, )))\n",
    "neural_network.add(Dense(16, activation='relu'))\n",
    "neural_network.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "# Compile the baseline network\n",
    "neural_network.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "# List a summary of the baseline network\n",
    "neural_network.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I fit the baseline network to the training data, and run it for 32 epochs alongside the validation data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/32\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.6045 - accuracy: 0.7016 - val_loss: 0.6316 - val_accuracy: 0.6778\n",
      "Epoch 2/32\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.5057 - accuracy: 0.7681 - val_loss: 0.5922 - val_accuracy: 0.7135\n",
      "Epoch 3/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4890 - accuracy: 0.7776 - val_loss: 0.5851 - val_accuracy: 0.7249\n",
      "Epoch 4/32\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4797 - accuracy: 0.7841 - val_loss: 0.6379 - val_accuracy: 0.6668\n",
      "Epoch 5/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4755 - accuracy: 0.7862 - val_loss: 0.5604 - val_accuracy: 0.7453\n",
      "Epoch 6/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4725 - accuracy: 0.7879 - val_loss: 0.5637 - val_accuracy: 0.7440\n",
      "Epoch 7/32\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4703 - accuracy: 0.7895 - val_loss: 0.5834 - val_accuracy: 0.7210\n",
      "Epoch 8/32\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4679 - accuracy: 0.7909 - val_loss: 0.5714 - val_accuracy: 0.7362\n",
      "Epoch 9/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4662 - accuracy: 0.7913 - val_loss: 0.5533 - val_accuracy: 0.7545\n",
      "Epoch 10/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4652 - accuracy: 0.7926 - val_loss: 0.5884 - val_accuracy: 0.7163\n",
      "Epoch 11/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4638 - accuracy: 0.7931 - val_loss: 0.5647 - val_accuracy: 0.7437\n",
      "Epoch 12/32\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4634 - accuracy: 0.7930 - val_loss: 0.5531 - val_accuracy: 0.7547\n",
      "Epoch 13/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4622 - accuracy: 0.7941 - val_loss: 0.5687 - val_accuracy: 0.7347\n",
      "Epoch 14/32\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4618 - accuracy: 0.7938 - val_loss: 0.5442 - val_accuracy: 0.7665\n",
      "Epoch 15/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4607 - accuracy: 0.7946 - val_loss: 0.5348 - val_accuracy: 0.7746\n",
      "Epoch 16/32\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4603 - accuracy: 0.7935 - val_loss: 0.5561 - val_accuracy: 0.7542\n",
      "Epoch 17/32\n",
      "2739/2739 [==============================] - 5s 2ms/step - loss: 0.4598 - accuracy: 0.7949 - val_loss: 0.5343 - val_accuracy: 0.7745\n",
      "Epoch 18/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4595 - accuracy: 0.7946 - val_loss: 0.5812 - val_accuracy: 0.7249\n",
      "Epoch 19/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4591 - accuracy: 0.7946 - val_loss: 0.5347 - val_accuracy: 0.7743\n",
      "Epoch 20/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4578 - accuracy: 0.7958 - val_loss: 0.5782 - val_accuracy: 0.7340\n",
      "Epoch 21/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4569 - accuracy: 0.7958 - val_loss: 0.5543 - val_accuracy: 0.7558\n",
      "Epoch 22/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4569 - accuracy: 0.7954 - val_loss: 0.5403 - val_accuracy: 0.7669\n",
      "Epoch 23/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4566 - accuracy: 0.7963 - val_loss: 0.5456 - val_accuracy: 0.7622\n",
      "Epoch 24/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4565 - accuracy: 0.7963 - val_loss: 0.5553 - val_accuracy: 0.7520\n",
      "Epoch 25/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4561 - accuracy: 0.7959 - val_loss: 0.5747 - val_accuracy: 0.7301\n",
      "Epoch 26/32\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4560 - accuracy: 0.7968 - val_loss: 0.5403 - val_accuracy: 0.7663\n",
      "Epoch 27/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4562 - accuracy: 0.7964 - val_loss: 0.5406 - val_accuracy: 0.7687\n",
      "Epoch 28/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4557 - accuracy: 0.7966 - val_loss: 0.5470 - val_accuracy: 0.7629\n",
      "Epoch 29/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4559 - accuracy: 0.7964 - val_loss: 0.5558 - val_accuracy: 0.7483\n",
      "Epoch 30/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4549 - accuracy: 0.7971 - val_loss: 0.5239 - val_accuracy: 0.7834\n",
      "Epoch 31/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4554 - accuracy: 0.7971 - val_loss: 0.5311 - val_accuracy: 0.7846\n",
      "Epoch 32/32\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4550 - accuracy: 0.7968 - val_loss: 0.5460 - val_accuracy: 0.7629\n"
     ]
    }
   ],
   "source": [
    "# Fit the baseline network\n",
    "baseline = neural_network.fit(X_train, y_train_res, epochs=32, batch_size=256,\n",
    "                                      validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Once all the epochs have been run through, I evaluate the network on both the training and validation data to get a look into the baseline networks accuracy and loss. Then I generate predictions and map the values to plot a confusion matrix, and get a further look into how the model performed. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21908/21908 [==============================] - 25s 1ms/step - loss: 36.8651 - accuracy: 0.5033\n",
      "Training data results:\n",
      "[36.86511993408203, 0.5033465027809143]\n",
      "\n",
      "\n",
      "2278/2278 [==============================] - 3s 1ms/step - loss: 0.5460 - accuracy: 0.7629\n",
      "Validation data results:\n",
      "[0.5460101962089539, 0.7629411816596985]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Evaluate the loss and accuracy scores for the training and validation datasets\n",
    "print(f'Training data results:\\n{neural_network.evaluate(X_train_res, y_train_res)}')\n",
    "print('\\n')\n",
    "print(f'Validation data results:\\n{neural_network.evaluate(X_val, y_val)}')\n",
    "\n",
    "# Generate predictions and \"round\" the values\n",
    "baseline_preds = neural_network.predict(X_val)\n",
    "baseline_preds[baseline_preds > 0.5] = 1\n",
    "baseline_preds[baseline_preds < 0.5] = 0\n",
    "\n",
    "# Plot a confusion matrix of the validation data\n",
    "plt.figure(figsize=(10, 8))\n",
    "sns.heatmap(confusion_matrix(y_val, baseline_preds), annot=True,\n",
    "           fmt='.1f', xticklabels=['Not Delayed', 'Delayed'],\n",
    "           yticklabels=['Not Delayed', 'Delayed'], cmap='OrRd', cbar=False)\n",
    "plt.xlabel(\"Predicted Labels\")\n",
    "plt.ylabel(\"True Labels\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I see above that the validation accuracy is much higher than the training data's score (76.3% vs. 50.3%, respectively). This is only a baseline network yet this performance is pretty poor and will need to be tuned and retrained. The confusion matrix above displays the model's performance, and gives us a better idea of how much work I need to do. When compared to the tuned machine learning model, the one above predicts more true positives (around 1,800 cases more) - essentially, more delay cases. Simultaneously, it predicts approximately 5,000 more false positives. If possible, I will try to greatly reduce that amount while also increasing the amount of true delays. \n",
    "\n",
    "Below is the classification report for the baseline network to see how the model performed based on the three metrics I've been tracking thus far (precision, recall and f1 score). However, for this analsysis and simplicity's sake, I also create a function further below that will take in the true and predicted labels, and generate the three scores."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.83      0.89      0.86     58461\n",
      "           1       0.36      0.27      0.31     14407\n",
      "\n",
      "    accuracy                           0.76     72868\n",
      "   macro avg       0.60      0.58      0.58     72868\n",
      "weighted avg       0.74      0.76      0.75     72868\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Print the classification report\n",
    "print(classification_report(y_val, baseline_preds))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision score: 36.37 %\n",
      "Recall score: 26.56 %\n",
      "F1 score: 30.7 %\n"
     ]
    }
   ],
   "source": [
    "# Create a function that will calculate the three relevant metric scores\n",
    "def model_metrics(y_true, y_preds):\n",
    "    \n",
    "    precision = precision_score(y_true, y_preds)\n",
    "    recall = recall_score(y_true, y_preds)\n",
    "    f1 = f1_score(y_true, y_preds)\n",
    "\n",
    "    print('Precision score:', round(precision * 100, 2),'%')\n",
    "    print('Recall score:', round(recall * 100, 2),'%')\n",
    "    print('F1 score:', round(f1 * 100, 2),'%')\n",
    "\n",
    "# Run the function with the appropriate arguments passed in\n",
    "model_metrics(y_val, baseline_preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It seems all three scores dropped when compared to the machine learning algorithm. The precision score is the only one to decrease when compared to the Random Forest classifier (36.4% vs 58.0%). Conversely, the recall and f1 scores increased in this iteration. This means the model's sensitivity, or ability to predict positive results, has improved, which we saw in our plot above. My goal, once again, is to improve this area while maintaining low false positive numbers.\n",
    "\n",
    "I will create another neural network; however, it will have 3 hidden layers, instead of two (and the units will be adjusted, accordingly). I apply the same compiler parameters, and print the summary again to see how many more parameters I am training. This denser network will train 11,249 total parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_3 (Dense)              (None, 64)                8896      \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 8)                 264       \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 1)                 9         \n",
      "=================================================================\n",
      "Total params: 11,249\n",
      "Trainable params: 11,249\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Create another, more dense neural network\n",
    "neural_network2 = Sequential()\n",
    "neural_network2.add(Dense(64, activation='relu', \n",
    "                            kernel_regularizer=regularizers.l2(0.01), input_shape=(138, )))\n",
    "neural_network2.add(Dense(32, activation='relu'))\n",
    "neural_network2.add(Dense(8, activation='relu'))\n",
    "neural_network2.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "# Compile the model\n",
    "neural_network2.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "# List a summary of the model\n",
    "neural_network2.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I fit the model this time to **50 epochs** in order to increase my training performance, and in the hope that my precision will increase as well. I will retain the batch size and evaluate on the validation data again."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.5701 - accuracy: 0.7285 - val_loss: 0.5611 - val_accuracy: 0.7535\n",
      "Epoch 2/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4853 - accuracy: 0.7784 - val_loss: 0.5793 - val_accuracy: 0.7297\n",
      "Epoch 3/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4726 - accuracy: 0.7868 - val_loss: 0.5876 - val_accuracy: 0.7219\n",
      "Epoch 4/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4668 - accuracy: 0.7909 - val_loss: 0.5436 - val_accuracy: 0.7685\n",
      "Epoch 5/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4626 - accuracy: 0.7931 - val_loss: 0.6136 - val_accuracy: 0.6953\n",
      "Epoch 6/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4602 - accuracy: 0.7950 - val_loss: 0.5713 - val_accuracy: 0.7383\n",
      "Epoch 7/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4580 - accuracy: 0.7962 - val_loss: 0.5497 - val_accuracy: 0.7608\n",
      "Epoch 8/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4566 - accuracy: 0.7972 - val_loss: 0.5285 - val_accuracy: 0.7895\n",
      "Epoch 9/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4553 - accuracy: 0.7981 - val_loss: 0.5474 - val_accuracy: 0.7631\n",
      "Epoch 10/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4556 - accuracy: 0.7980 - val_loss: 0.5258 - val_accuracy: 0.7864\n",
      "Epoch 11/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4548 - accuracy: 0.7985 - val_loss: 0.5811 - val_accuracy: 0.7272\n",
      "Epoch 12/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4553 - accuracy: 0.7985 - val_loss: 0.5535 - val_accuracy: 0.7582\n",
      "Epoch 13/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4539 - accuracy: 0.7993 - val_loss: 0.5334 - val_accuracy: 0.7760\n",
      "Epoch 14/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4545 - accuracy: 0.7989 - val_loss: 0.5435 - val_accuracy: 0.7609\n",
      "Epoch 15/50\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.4545 - accuracy: 0.7992 - val_loss: 0.5680 - val_accuracy: 0.7390\n",
      "Epoch 16/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4544 - accuracy: 0.7990 - val_loss: 0.5364 - val_accuracy: 0.7735\n",
      "Epoch 17/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4538 - accuracy: 0.7990 - val_loss: 0.5463 - val_accuracy: 0.7632\n",
      "Epoch 18/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4541 - accuracy: 0.7986 - val_loss: 0.5344 - val_accuracy: 0.7783\n",
      "Epoch 19/50\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.4542 - accuracy: 0.7989 - val_loss: 0.5512 - val_accuracy: 0.7614\n",
      "Epoch 20/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4534 - accuracy: 0.7996 - val_loss: 0.5464 - val_accuracy: 0.7704\n",
      "Epoch 21/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4528 - accuracy: 0.7999 - val_loss: 0.5882 - val_accuracy: 0.7240\n",
      "Epoch 22/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4530 - accuracy: 0.7995 - val_loss: 0.5522 - val_accuracy: 0.7600\n",
      "Epoch 23/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4523 - accuracy: 0.8000 - val_loss: 0.5335 - val_accuracy: 0.7800\n",
      "Epoch 24/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4514 - accuracy: 0.8000 - val_loss: 0.5412 - val_accuracy: 0.7707\n",
      "Epoch 25/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4511 - accuracy: 0.7998 - val_loss: 0.5287 - val_accuracy: 0.7838\n",
      "Epoch 26/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4502 - accuracy: 0.8010 - val_loss: 0.5434 - val_accuracy: 0.7692\n",
      "Epoch 27/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4494 - accuracy: 0.8011 - val_loss: 0.5582 - val_accuracy: 0.7496\n",
      "Epoch 28/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4487 - accuracy: 0.8014 - val_loss: 0.5422 - val_accuracy: 0.7690\n",
      "Epoch 29/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4489 - accuracy: 0.8012 - val_loss: 0.5526 - val_accuracy: 0.7561\n",
      "Epoch 30/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4473 - accuracy: 0.8013 - val_loss: 0.5523 - val_accuracy: 0.7539\n",
      "Epoch 31/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4460 - accuracy: 0.8022 - val_loss: 0.5491 - val_accuracy: 0.7565\n",
      "Epoch 32/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4432 - accuracy: 0.8034 - val_loss: 0.5702 - val_accuracy: 0.7426\n",
      "Epoch 33/50\n",
      "2739/2739 [==============================] - 12s 4ms/step - loss: 0.4418 - accuracy: 0.8029 - val_loss: 0.5341 - val_accuracy: 0.7797\n",
      "Epoch 34/50\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.4413 - accuracy: 0.8034 - val_loss: 0.5349 - val_accuracy: 0.7752\n",
      "Epoch 35/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4405 - accuracy: 0.8033 - val_loss: 0.5648 - val_accuracy: 0.7525\n",
      "Epoch 36/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4405 - accuracy: 0.8028 - val_loss: 0.5266 - val_accuracy: 0.7893\n",
      "Epoch 37/50\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.4400 - accuracy: 0.8024 - val_loss: 0.5210 - val_accuracy: 0.7922\n",
      "Epoch 38/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4401 - accuracy: 0.8020 - val_loss: 0.5538 - val_accuracy: 0.7645\n",
      "Epoch 39/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4402 - accuracy: 0.8019 - val_loss: 0.5356 - val_accuracy: 0.7808\n",
      "Epoch 40/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4395 - accuracy: 0.8019 - val_loss: 0.5797 - val_accuracy: 0.7464\n",
      "Epoch 41/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4401 - accuracy: 0.8018 - val_loss: 0.5623 - val_accuracy: 0.7665\n",
      "Epoch 42/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4396 - accuracy: 0.8023 - val_loss: 0.5220 - val_accuracy: 0.7918\n",
      "Epoch 43/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4396 - accuracy: 0.8021 - val_loss: 0.5403 - val_accuracy: 0.7738\n",
      "Epoch 44/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4394 - accuracy: 0.8023 - val_loss: 0.5590 - val_accuracy: 0.7608\n",
      "Epoch 45/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4386 - accuracy: 0.8025 - val_loss: 0.5573 - val_accuracy: 0.7653\n",
      "Epoch 46/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4387 - accuracy: 0.8015 - val_loss: 0.5864 - val_accuracy: 0.7341\n",
      "Epoch 47/50\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4388 - accuracy: 0.8024 - val_loss: 0.5273 - val_accuracy: 0.7834\n",
      "Epoch 48/50\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.4387 - accuracy: 0.8019 - val_loss: 0.5383 - val_accuracy: 0.7739\n",
      "Epoch 49/50\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4389 - accuracy: 0.8019 - val_loss: 0.5316 - val_accuracy: 0.7796\n",
      "Epoch 50/50\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4386 - accuracy: 0.8026 - val_loss: 0.5223 - val_accuracy: 0.7874\n"
     ]
    }
   ],
   "source": [
    "# Fit the model\n",
    "results = neural_network2.fit(X_train, y_train_res, epochs=50, batch_size=256,\n",
    "                                      validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    1/21908 [..............................] - ETA: 0s - loss: 2.3469 - accuracy: 0.8750WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0010s vs `on_test_batch_end` time: 0.0018s). Check your callbacks.\n",
      "21908/21908 [==============================] - 27s 1ms/step - loss: 14.5650 - accuracy: 0.5102\n",
      "Training data results:\n",
      "[14.565011978149414, 0.5101722478866577]\n",
      "\n",
      "\n",
      "2278/2278 [==============================] - 3s 1ms/step - loss: 0.5223 - accuracy: 0.7874\n",
      "Validation data results:\n",
      "[0.5222695469856262, 0.787423849105835]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Evaluate the loss and accuracy scores for the training and validation datasets\n",
    "print(f'Training data results:\\n{neural_network2.evaluate(X_train_res, y_train_res)}')\n",
    "print('\\n')\n",
    "print(f'Validation data results:\\n{neural_network2.evaluate(X_val, y_val)}')\n",
    "\n",
    "# Generate predictions and \"round\" the values\n",
    "y_preds2 = neural_network2.predict(X_val)\n",
    "y_preds2[y_preds2 > 0.5] = 1\n",
    "y_preds2[y_preds2 < 0.5] = 0\n",
    "\n",
    "# Plot a confusion matrix of the validation data\n",
    "plt.figure(figsize=(10, 8))\n",
    "sns.heatmap(confusion_matrix(y_val, y_preds2), annot=True,\n",
    "           fmt='.1f', xticklabels=['Not Delayed', 'Delayed'],\n",
    "           yticklabels=['Not Delayed', 'Delayed'], cmap='OrRd', cbar=False)\n",
    "plt.xlabel(\"Predicted Labels\")\n",
    "plt.ylabel(\"True Labels\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Based on the training and validation accuracies, I see that my model got better at misclassifying delayed flights! The true delays in my confusion matrix though got worse. My false positive cases reduced from 6,694 to 3,105, which is a good start towards my goals. However, my delay cases decreased slightly and needs to be improved. I print out the three metrics once again to get a better idea of my model's performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision score: 39.44 %\n",
      "Recall score: 14.03 %\n",
      "F1 score: 20.7 %\n"
     ]
    }
   ],
   "source": [
    "# Run the metrics function\n",
    "model_metrics(y_val, y_preds2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The three scores above tell me a few things. First, the precision score slightly increased to 39.4% and the remaining two scores decreased, which is to be expected after seeing the true positive cases decrease. I will create one more neural network to try to improve my model's performance once again. Below I create another network, reverting back to my baseline architecture but with a smaller regularizer (0.001)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_7 (Dense)              (None, 64)                8896      \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 16)                1040      \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1)                 17        \n",
      "=================================================================\n",
      "Total params: 9,953\n",
      "Trainable params: 9,953\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Create a neural network similar to the baseline with a smaller regularizer\n",
    "neural_network3 = Sequential()\n",
    "neural_network3.add(Dense(64, activation='relu', \n",
    "                            kernel_regularizer=regularizers.l2(0.001), input_shape=(138, )))\n",
    "neural_network3.add(Dense(16, activation='relu'))\n",
    "neural_network3.add(Dense(1, activation='sigmoid'))\n",
    "\n",
    "# Compile the baseline model\n",
    "neural_network3.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "\n",
    "# List a summary of the model\n",
    "neural_network3.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This iteration will differ from the other two networks I built by increasing the epochs to 128, setting a stopping condition with the `EarlyStopping` function, and saving that checkpoint as my best deep learning model with the `ModelCheckpoint` function. I will pass this stopping condition as I fit my model, evaluating it once again on my validation data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a file name object for the best model checkpoint\n",
    "model_filepath = 'deep_learning_model.h5'\n",
    "\n",
    "# Set early stopping and model checkpoint conditions\n",
    "early_stopping = [EarlyStopping(monitor='val_loss', patience=8), \n",
    "                  ModelCheckpoint(filepath=model_filepath, monitor='val_loss', save_best_only=True)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/128\n",
      "2739/2739 [==============================] - 8s 3ms/step - loss: 0.5449 - accuracy: 0.7335 - val_loss: 0.6111 - val_accuracy: 0.6918\n",
      "Epoch 2/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4556 - accuracy: 0.7943 - val_loss: 0.5701 - val_accuracy: 0.7268\n",
      "Epoch 3/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4389 - accuracy: 0.8036 - val_loss: 0.5330 - val_accuracy: 0.7650\n",
      "Epoch 4/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4315 - accuracy: 0.8079 - val_loss: 0.5623 - val_accuracy: 0.7351\n",
      "Epoch 5/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4264 - accuracy: 0.8107 - val_loss: 0.5628 - val_accuracy: 0.7325\n",
      "Epoch 6/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4239 - accuracy: 0.8122 - val_loss: 0.5128 - val_accuracy: 0.7917\n",
      "Epoch 7/128\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4216 - accuracy: 0.8134 - val_loss: 0.5598 - val_accuracy: 0.7368\n",
      "Epoch 8/128\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4197 - accuracy: 0.8145 - val_loss: 0.5432 - val_accuracy: 0.7543\n",
      "Epoch 9/128\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4185 - accuracy: 0.8151 - val_loss: 0.5527 - val_accuracy: 0.7407\n",
      "Epoch 10/128\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4165 - accuracy: 0.8168 - val_loss: 0.5262 - val_accuracy: 0.7715\n",
      "Epoch 11/128\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4157 - accuracy: 0.8163 - val_loss: 0.5350 - val_accuracy: 0.7606\n",
      "Epoch 12/128\n",
      "2739/2739 [==============================] - 7s 3ms/step - loss: 0.4139 - accuracy: 0.8178 - val_loss: 0.5448 - val_accuracy: 0.7539\n",
      "Epoch 13/128\n",
      "2739/2739 [==============================] - 7s 2ms/step - loss: 0.4135 - accuracy: 0.8182 - val_loss: 0.5309 - val_accuracy: 0.7667\n",
      "Epoch 14/128\n",
      "2739/2739 [==============================] - 6s 2ms/step - loss: 0.4122 - accuracy: 0.8191 - val_loss: 0.5237 - val_accuracy: 0.7712\n"
     ]
    }
   ],
   "source": [
    "# Fit the model with more epochs\n",
    "results2 = neural_network3.fit(X_train, y_train_res, epochs=128, batch_size=256, callbacks=early_stopping,\n",
    "                                      validation_data=(X_val, y_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I see that my model has stopped after 14 epochs! Now I will take that saved model, load it, and evaluate the model on the training and validation data to view its accuracy. I will also plot the confusion matrix, and see how the model performed. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21908/21908 [==============================] - 27s 1ms/step - loss: 0.4338 - accuracy: 0.8143\n",
      "Training Loss: 0.434 \n",
      "Training Accuracy: 0.814\n",
      "----------\n",
      "2278/2278 [==============================] - 3s 1ms/step - loss: 0.5128 - accuracy: 0.7917: \n",
      "Validation Loss: 0.513 \n",
      "Validation Accuracy: 0.792\n"
     ]
    }
   ],
   "source": [
    "# Load the saved model\n",
    "saved_model = load_model(model_filepath)\n",
    "\n",
    "# Calculate the loss and accuracy scores for both training datasets\n",
    "results_train = saved_model.evaluate(X_train, y_train_res)\n",
    "print(f'Training Loss: {results_train[0]:.3} \\nTraining Accuracy: {results_train[1]:.3}')\n",
    "\n",
    "print('----------')\n",
    "\n",
    "results_val = saved_model.evaluate(X_val, y_val)\n",
    "print(f'Validation Loss: {results_val[0]:.3} \\nValidation Accuracy: {results_val[1]:.3}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Generate predictions and \"round\" the values\n",
    "y_val_preds = saved_model.predict(X_val)\n",
    "y_val_preds[y_val_preds > 0.5] = 1\n",
    "y_val_preds[y_val_preds < 0.5] = 0\n",
    "\n",
    "# Plot a confusion matrix of the validation data\n",
    "plt.figure(figsize=(10, 8))\n",
    "sns.heatmap(confusion_matrix(y_val, y_val_preds), annot=True,\n",
    "           fmt='.1f', xticklabels=['Not Delayed', 'Delayed'],\n",
    "           yticklabels=['Not Delayed', 'Delayed'], cmap='OrRd', cbar=False)\n",
    "plt.xlabel(\"Predicted Labels\")\n",
    "plt.ylabel(\"True Labels\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is definitely the best deep learning model at precisely avoid miscategorizing non-delayed flights; although, the other metrics leave something to be desired. I can see that the false positive cases amount to 2,786, which makes this good enough to serve as my best deep learning model - and is also saved. By calculating the model's scores, I also see that the model is 42.0% precise, has a recall of 14.0%, and its f1 score is at 21.0%. The latter metrics are exhibited in the plot above by the decrease in the true delays predictions. As a matter of fact, the 12,389 cases of non-delayed predictions is the highest I've seen yet, showing our model isn't getting better in that regard. There will need to be more work done on the network's architecture to better improve that area, the extent to which will also depend on resource constraints. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision score: 42.01 %\n",
      "Recall score: 14.01 %\n",
      "F1 score: 21.01 %\n"
     ]
    }
   ],
   "source": [
    "# Calculate the relevant metrics\n",
    "model_metrics(y_val, y_val_preds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluation\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When comparing models, I decide that the machine learning technique produced the best one, and should be used to evaluate my test data. I will generate the predictions, and print the three scores to get a full picture of my model's efficiency. Fortunately, the model has been already saved and has been included in the repository as my means of deployment (with improved resources, the aim is to deploy the model as an app for daily use). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the ML model  \n",
    "with open('best_model.pkl', 'wb') as f:\n",
    "    joblib.dump(rf_gridsearch, f)\n",
    "\n",
    "best_model = joblib.load('best_model.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generated 72867 predictions\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      " Not Delayed       0.81      0.90      0.85     58405\n",
      "     Delayed       0.27      0.16      0.20     14462\n",
      "\n",
      "    accuracy                           0.75     72867\n",
      "   macro avg       0.54      0.53      0.53     72867\n",
      "weighted avg       0.70      0.75      0.72     72867\n",
      "\n",
      "Precision score: 27.15 %\n",
      "Recall score: 15.79 %\n",
      "F1 score: 19.97 %\n"
     ]
    }
   ],
   "source": [
    "# Evaluate the test data and list the relevant metrics\n",
    "y_hat_test = best_model.predict(X_test)\n",
    "\n",
    "target_names = ['Not Delayed', 'Delayed']\n",
    "print('Generated {} predictions'.format(len(y_hat_test)))\n",
    "print(classification_report(y_test, y_hat_test, target_names=target_names))\n",
    "model_metrics(y_test, y_hat_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "My model's test metrics are worse than when it performed on my validation data. Specifically, the model is 27% precise, has a recall of 16.0%, and an f1 score of 20.0%. This is not good and the model needs to be refined. There are some limitations I need to improve, which I will address after this evaluation, that I'll need to overcome to deploy this model for public use. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot a confusion matrix of the test data\n",
    "plt.figure(figsize=(10, 8))\n",
    "sns.heatmap(confusion_matrix(y_test, y_hat_test), annot=True,\n",
    "           fmt='.1f', xticklabels=['Not Delayed', 'Delayed'],\n",
    "           yticklabels=['Not Delayed', 'Delayed'], cmap='OrRd', cbar=False)\n",
    "plt.xlabel(\"Predicted Labels\")\n",
    "plt.ylabel(\"True Labels\")\n",
    "plt.show();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The confusion matrix above only further displays the model's test performance - there are 6,127 false positive predictions, which is worse than the past models I built and evaluated. The number of true positive cases increased, which gives me some hope that the model can be improved to levels that are acceptable for mass deployment. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This analysis leads to the following conclusions:\n",
    "\n",
    "1. The machine learning model performed best, and is the analysis technique I will further apply to improve my performance.\n",
    "\n",
    "2. The model is exactly 27.2% precise when correctly classifying delayed flights."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Limitations/Further Work"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This project is limited in a few ways. First and foremost, I built my models under heavy computational constraints. Given the nature of the data, it is necessary to train models on computers that can process larger and fuller datasets in quicker time. For example, the grid search I performed earlier took almost 3 hours to run, despite a thin parameter grid. With improved hardware, I could use that same amount of time but apply techniques, like creating a better (more complete) parameter grid or analyzing larger data, to improve the model's performance. Another drawback is the class imbalance. I applied sampling methods to augment the minority class, which gave me more data to work with, but also increased my run times. Therefore, gathering more data that fall in the minority class would greatly improve my precision. Lastly, and this ties with my computational constraints, I mentioned I could have used a wider range of hyperparameters to perform my grid search with. Unfortunately, with limited resources, it would have taken me hours or even days to fully perform this search. However, I could have found a better set of hyperparameters that would improve my metrics and predictions. \n",
    "\n",
    "Further analyses could yield a more effective model, and possibly improve the algorithm's performance to the point where I push to market. Some possible courses of action I could take include:\n",
    "\n",
    "1. Training my model with better, stronger hardware/computers.\n",
    "2. Gathering more data with an emphasis on balancing the minority class to avoid sampling. \n",
    "3. Evaluating on different parameters to see how it impacts my model's performance.\n",
    "4. Gain more feedback from peers and analyze the model with other methods (i.e., the Gradient Boosting classifier)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sources\n",
    "Link to original dataset: \n",
    "https://www.kaggle.com/datasets/threnjen/2019-airline-delays-and-cancellations/data?select=full_data_flightdelay.csv\n",
    "\n",
    "Wall Street Journal annual airlines ranking report:\n",
    "https://www.wsj.com/lifestyle/travel/best-airlines-us-2023-36e9ea20\n",
    "\n",
    "Bureau of Transportation Statistics:\n",
    "https://www.transtats.bts.gov/HomeDrillChart.asp\n",
    "\n",
    "Federal Aviation Administration (FAA):\n",
    "https://aspm.faa.gov/aspmhelp/index/Types_of_Delay.html"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (learn-env",
   "language": "python",
   "name": "learn-env"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}