Predicting flight delays.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predicting Flight Delay\n",
    "\n",
    "**Problem Set-up:**\n",
    "We define a delayed flight to be one that is delayed by >= 15 minutes. \n",
    "The prediction problem is to train a model that can classify flights, to predict if they will or will not be delayed.\n",
    "\n",
    "**Use case:**\n",
    "The idea is that this model would be useful to choosing airlines, flightpaths, airports, at the time of booking, relatively in advance of the scheduled departure (days, weeks, months ahead of time). Therefore, the prediction problem will focus on features that can be known in advance, rather than predicting using day-off features like weather and previous flights from that day. \n",
    "\n",
    "**Approach:**\n",
    "Train a logistic regression using one-hot encoding of features explored in the accompanying notebook \"trends in flight delays.\" Create separate models to predict for each airport."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Imports\n",
    "%config InlineBackend.figure_format = 'retina'\n",
    "%matplotlib inline\n",
    "\n",
    "import numpy as np\n",
    "import scipy as sp\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "\n",
    "import seaborn as sns\n",
    "sns.set_style('white')\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn. metrics import roc_auc_score, confusion_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 0. Load and prep data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0a. Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Import custom code\n",
    "from flightdelay.fld import io as flio\n",
    "airlines_df, airports_df, flights_df = flio.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Filter data to keys of interest\n",
    "keys = ['MONTH', 'DAY', 'DAY_OF_WEEK', 'AIRLINE', 'FLIGHT_NUMBER', 'TAIL_NUMBER', 'ORIGIN_AIRPORT',\n",
    "       'DESTINATION_AIRPORT', 'SCHEDULED_DEPARTURE', 'DEPARTURE_TIME',\n",
    "       'DEPARTURE_DELAY', 'TAXI_OUT', 'WHEELS_OFF', 'SCHEDULED_TIME',\n",
    "       'ELAPSED_TIME', 'AIR_TIME', 'DISTANCE', 'WHEELS_ON', 'TAXI_IN',\n",
    "       'SCHEDULED_ARRIVAL', 'ARRIVAL_TIME', 'ARRIVAL_DELAY']\n",
    "flights_df = flights_df[keys]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0b.  Remove airports with less than a certain number of flights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "all_airports, airport_inverse, airport_count = np.unique(flights_df['ORIGIN_AIRPORT'],return_counts=True,return_inverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Determine number of flights for the origin airport\n",
    "Nflights_orig = np.zeros(len(airport_inverse))\n",
    "for i in range(len(all_airports)):\n",
    "    Nflights_orig[np.where(airport_inverse==i)] = airport_count[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "flights_df = flights_df.loc[flights_df.index[Nflights_orig>=7300]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0c. Remove cancelled flights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MONTH</th>\n",
       "      <th>DAY</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>AIRLINE</th>\n",
       "      <th>FLIGHT_NUMBER</th>\n",
       "      <th>TAIL_NUMBER</th>\n",
       "      <th>ORIGIN_AIRPORT</th>\n",
       "      <th>DESTINATION_AIRPORT</th>\n",
       "      <th>SCHEDULED_DEPARTURE</th>\n",
       "      <th>DEPARTURE_TIME</th>\n",
       "      <th>...</th>\n",
       "      <th>WHEELS_OFF</th>\n",
       "      <th>SCHEDULED_TIME</th>\n",
       "      <th>ELAPSED_TIME</th>\n",
       "      <th>AIR_TIME</th>\n",
       "      <th>DISTANCE</th>\n",
       "      <th>WHEELS_ON</th>\n",
       "      <th>TAXI_IN</th>\n",
       "      <th>SCHEDULED_ARRIVAL</th>\n",
       "      <th>ARRIVAL_TIME</th>\n",
       "      <th>ARRIVAL_DELAY</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>AS</td>\n",
       "      <td>98</td>\n",
       "      <td>N407AS</td>\n",
       "      <td>ANC</td>\n",
       "      <td>SEA</td>\n",
       "      <td>5</td>\n",
       "      <td>2354.0</td>\n",
       "      <td>...</td>\n",
       "      <td>15.0</td>\n",
       "      <td>205.0</td>\n",
       "      <td>194.0</td>\n",
       "      <td>169.0</td>\n",
       "      <td>1448</td>\n",
       "      <td>404.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>430</td>\n",
       "      <td>408.0</td>\n",
       "      <td>-22.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>AA</td>\n",
       "      <td>2336</td>\n",
       "      <td>N3KUAA</td>\n",
       "      <td>LAX</td>\n",
       "      <td>PBI</td>\n",
       "      <td>10</td>\n",
       "      <td>2.0</td>\n",
       "      <td>...</td>\n",
       "      <td>14.0</td>\n",
       "      <td>280.0</td>\n",
       "      <td>279.0</td>\n",
       "      <td>263.0</td>\n",
       "      <td>2330</td>\n",
       "      <td>737.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>750</td>\n",
       "      <td>741.0</td>\n",
       "      <td>-9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>US</td>\n",
       "      <td>840</td>\n",
       "      <td>N171US</td>\n",
       "      <td>SFO</td>\n",
       "      <td>CLT</td>\n",
       "      <td>20</td>\n",
       "      <td>18.0</td>\n",
       "      <td>...</td>\n",
       "      <td>34.0</td>\n",
       "      <td>286.0</td>\n",
       "      <td>293.0</td>\n",
       "      <td>266.0</td>\n",
       "      <td>2296</td>\n",
       "      <td>800.0</td>\n",
       "      <td>11.0</td>\n",
       "      <td>806</td>\n",
       "      <td>811.0</td>\n",
       "      <td>5.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>AA</td>\n",
       "      <td>258</td>\n",
       "      <td>N3HYAA</td>\n",
       "      <td>LAX</td>\n",
       "      <td>MIA</td>\n",
       "      <td>20</td>\n",
       "      <td>15.0</td>\n",
       "      <td>...</td>\n",
       "      <td>30.0</td>\n",
       "      <td>285.0</td>\n",
       "      <td>281.0</td>\n",
       "      <td>258.0</td>\n",
       "      <td>2342</td>\n",
       "      <td>748.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>805</td>\n",
       "      <td>756.0</td>\n",
       "      <td>-9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>AS</td>\n",
       "      <td>135</td>\n",
       "      <td>N527AS</td>\n",
       "      <td>SEA</td>\n",
       "      <td>ANC</td>\n",
       "      <td>25</td>\n",
       "      <td>24.0</td>\n",
       "      <td>...</td>\n",
       "      <td>35.0</td>\n",
       "      <td>235.0</td>\n",
       "      <td>215.0</td>\n",
       "      <td>199.0</td>\n",
       "      <td>1448</td>\n",
       "      <td>254.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>320</td>\n",
       "      <td>259.0</td>\n",
       "      <td>-21.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   MONTH  DAY  DAY_OF_WEEK AIRLINE  FLIGHT_NUMBER TAIL_NUMBER ORIGIN_AIRPORT  \\\n",
       "0      1    1            4      AS             98      N407AS            ANC   \n",
       "1      1    1            4      AA           2336      N3KUAA            LAX   \n",
       "2      1    1            4      US            840      N171US            SFO   \n",
       "3      1    1            4      AA            258      N3HYAA            LAX   \n",
       "4      1    1            4      AS            135      N527AS            SEA   \n",
       "\n",
       "  DESTINATION_AIRPORT  SCHEDULED_DEPARTURE  DEPARTURE_TIME      ...        \\\n",
       "0                 SEA                    5          2354.0      ...         \n",
       "1                 PBI                   10             2.0      ...         \n",
       "2                 CLT                   20            18.0      ...         \n",
       "3                 MIA                   20            15.0      ...         \n",
       "4                 ANC                   25            24.0      ...         \n",
       "\n",
       "   WHEELS_OFF  SCHEDULED_TIME  ELAPSED_TIME  AIR_TIME  DISTANCE  WHEELS_ON  \\\n",
       "0        15.0           205.0         194.0     169.0      1448      404.0   \n",
       "1        14.0           280.0         279.0     263.0      2330      737.0   \n",
       "2        34.0           286.0         293.0     266.0      2296      800.0   \n",
       "3        30.0           285.0         281.0     258.0      2342      748.0   \n",
       "4        35.0           235.0         215.0     199.0      1448      254.0   \n",
       "\n",
       "   TAXI_IN  SCHEDULED_ARRIVAL  ARRIVAL_TIME  ARRIVAL_DELAY  \n",
       "0      4.0                430         408.0          -22.0  \n",
       "1      4.0                750         741.0           -9.0  \n",
       "2     11.0                806         811.0            5.0  \n",
       "3      8.0                805         756.0           -9.0  \n",
       "4      5.0                320         259.0          -21.0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "flights_df = flights_df.dropna()\n",
    "flights_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0d. Shuffle data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>MONTH</th>\n",
       "      <th>DAY</th>\n",
       "      <th>DAY_OF_WEEK</th>\n",
       "      <th>AIRLINE</th>\n",
       "      <th>FLIGHT_NUMBER</th>\n",
       "      <th>TAIL_NUMBER</th>\n",
       "      <th>ORIGIN_AIRPORT</th>\n",
       "      <th>DESTINATION_AIRPORT</th>\n",
       "      <th>SCHEDULED_DEPARTURE</th>\n",
       "      <th>DEPARTURE_TIME</th>\n",
       "      <th>...</th>\n",
       "      <th>WHEELS_OFF</th>\n",
       "      <th>SCHEDULED_TIME</th>\n",
       "      <th>ELAPSED_TIME</th>\n",
       "      <th>AIR_TIME</th>\n",
       "      <th>DISTANCE</th>\n",
       "      <th>WHEELS_ON</th>\n",
       "      <th>TAXI_IN</th>\n",
       "      <th>SCHEDULED_ARRIVAL</th>\n",
       "      <th>ARRIVAL_TIME</th>\n",
       "      <th>ARRIVAL_DELAY</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8</td>\n",
       "      <td>23</td>\n",
       "      <td>7</td>\n",
       "      <td>WN</td>\n",
       "      <td>285</td>\n",
       "      <td>N7738A</td>\n",
       "      <td>MCI</td>\n",
       "      <td>DEN</td>\n",
       "      <td>1410</td>\n",
       "      <td>1408.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1414.0</td>\n",
       "      <td>105.0</td>\n",
       "      <td>95.0</td>\n",
       "      <td>83.0</td>\n",
       "      <td>533</td>\n",
       "      <td>1437.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>1455</td>\n",
       "      <td>1443.0</td>\n",
       "      <td>-12.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>12</td>\n",
       "      <td>5</td>\n",
       "      <td>6</td>\n",
       "      <td>AS</td>\n",
       "      <td>335</td>\n",
       "      <td>N520AS</td>\n",
       "      <td>SJC</td>\n",
       "      <td>SEA</td>\n",
       "      <td>1550</td>\n",
       "      <td>1534.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1554.0</td>\n",
       "      <td>120.0</td>\n",
       "      <td>135.0</td>\n",
       "      <td>105.0</td>\n",
       "      <td>697</td>\n",
       "      <td>1739.0</td>\n",
       "      <td>10.0</td>\n",
       "      <td>1750</td>\n",
       "      <td>1749.0</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>12</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>UA</td>\n",
       "      <td>1750</td>\n",
       "      <td>N76254</td>\n",
       "      <td>SAN</td>\n",
       "      <td>IAH</td>\n",
       "      <td>1314</td>\n",
       "      <td>1354.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1406.0</td>\n",
       "      <td>188.0</td>\n",
       "      <td>168.0</td>\n",
       "      <td>148.0</td>\n",
       "      <td>1303</td>\n",
       "      <td>1834.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>1822</td>\n",
       "      <td>1842.0</td>\n",
       "      <td>20.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5</td>\n",
       "      <td>31</td>\n",
       "      <td>7</td>\n",
       "      <td>WN</td>\n",
       "      <td>396</td>\n",
       "      <td>N8318F</td>\n",
       "      <td>AUS</td>\n",
       "      <td>DEN</td>\n",
       "      <td>1935</td>\n",
       "      <td>2039.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2051.0</td>\n",
       "      <td>140.0</td>\n",
       "      <td>128.0</td>\n",
       "      <td>109.0</td>\n",
       "      <td>775</td>\n",
       "      <td>2140.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>2055</td>\n",
       "      <td>2147.0</td>\n",
       "      <td>52.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>26</td>\n",
       "      <td>5</td>\n",
       "      <td>EV</td>\n",
       "      <td>4899</td>\n",
       "      <td>N133EV</td>\n",
       "      <td>ATL</td>\n",
       "      <td>MYR</td>\n",
       "      <td>1454</td>\n",
       "      <td>1509.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1533.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>77.0</td>\n",
       "      <td>49.0</td>\n",
       "      <td>317</td>\n",
       "      <td>1622.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>1619</td>\n",
       "      <td>1626.0</td>\n",
       "      <td>7.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   MONTH  DAY  DAY_OF_WEEK AIRLINE  FLIGHT_NUMBER TAIL_NUMBER ORIGIN_AIRPORT  \\\n",
       "0      8   23            7      WN            285      N7738A            MCI   \n",
       "1     12    5            6      AS            335      N520AS            SJC   \n",
       "2     12    1            2      UA           1750      N76254            SAN   \n",
       "3      5   31            7      WN            396      N8318F            AUS   \n",
       "4      6   26            5      EV           4899      N133EV            ATL   \n",
       "\n",
       "  DESTINATION_AIRPORT  SCHEDULED_DEPARTURE  DEPARTURE_TIME      ...        \\\n",
       "0                 DEN                 1410          1408.0      ...         \n",
       "1                 SEA                 1550          1534.0      ...         \n",
       "2                 IAH                 1314          1354.0      ...         \n",
       "3                 DEN                 1935          2039.0      ...         \n",
       "4                 MYR                 1454          1509.0      ...         \n",
       "\n",
       "   WHEELS_OFF  SCHEDULED_TIME  ELAPSED_TIME  AIR_TIME  DISTANCE  WHEELS_ON  \\\n",
       "0      1414.0           105.0          95.0      83.0       533     1437.0   \n",
       "1      1554.0           120.0         135.0     105.0       697     1739.0   \n",
       "2      1406.0           188.0         168.0     148.0      1303     1834.0   \n",
       "3      2051.0           140.0         128.0     109.0       775     2140.0   \n",
       "4      1533.0            85.0          77.0      49.0       317     1622.0   \n",
       "\n",
       "   TAXI_IN  SCHEDULED_ARRIVAL  ARRIVAL_TIME  ARRIVAL_DELAY  \n",
       "0      6.0               1455        1443.0          -12.0  \n",
       "1     10.0               1750        1749.0           -1.0  \n",
       "2      8.0               1822        1842.0           20.0  \n",
       "3      7.0               2055        2147.0           52.0  \n",
       "4      4.0               1619        1626.0            7.0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shuffle data\n",
    "delay_cutoff = 15\n",
    "\n",
    "flights_df = flights_df.reset_index(drop=True)\n",
    "labels_preshuffle = flights_df['DEPARTURE_DELAY'].values > delay_cutoff\n",
    "\n",
    "np.random.seed(0)\n",
    "flight_shuff_idx = np.random.permutation(flights_df.index)\n",
    "labels_shuffle = labels_preshuffle[flight_shuff_idx]\n",
    "\n",
    "flights_df = flights_df.loc[flight_shuff_idx]\n",
    "\n",
    "flights_df = flights_df.reset_index(drop=True)\n",
    "flights_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Compute features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# General functions for computing features\n",
    "def make_onehot_feat_dict_from_vals(df, feat_key, feat_name, feat_cutoffs):\n",
    "    \n",
    "    # Create keys\n",
    "    N_feat = len(feat_cutoffs) - 1\n",
    "    keys = [0]*N_feat\n",
    "    for i in range(N_feat):\n",
    "        keys[i] = 'f_'+feat_name+'_'+ str(feat_cutoffs[i])\n",
    "    # Find the indices for each class\n",
    "    feat_dict = {}\n",
    "    for i in range(N_feat):\n",
    "        feat_dict[keys[i]] = np.transpose(\n",
    "                    np.logical_and(df[feat_key].values>= feat_cutoffs[i],\n",
    "                                df[feat_key].values<feat_cutoffs[i+1]))\n",
    "    return feat_dict\n",
    "\n",
    "\n",
    "def make_onehot_feat_dict(df, feat_key, feat_name):\n",
    "    # Create features for each day of the week\n",
    "    feat_vals = df[feat_key].values\n",
    "    all_vals = np.unique(feat_vals)\n",
    "    N_vals = len(all_vals)\n",
    "    N_feat = N_vals - 1\n",
    "\n",
    "    # Create keys\n",
    "    keys = [0]*N_feat\n",
    "    for i in range(N_feat):\n",
    "        keys[i] = 'f_'+feat_name+'_'+ str(all_vals[i])\n",
    "\n",
    "    # Create value for each training example in dict\n",
    "    feat_dict = {}\n",
    "    for i, k in enumerate(keys):\n",
    "        this_day = all_vals[i]\n",
    "        feat_dict[k] = feat_vals == this_day\n",
    "    return feat_dict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1a. Duration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "cutoffs = np.arange(60,780,60)\n",
    "original_feat = make_onehot_feat_dict_from_vals(flights_df,\n",
    "                    'SCHEDULED_TIME', 'dur', cutoffs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1b. Day of week"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "daysfeat_dict = make_onehot_feat_dict(flights_df, 'DAY_OF_WEEK', 'day')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1c. Month"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "monthsfeat_dict = make_onehot_feat_dict(flights_df, 'MONTH', 'month') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1d. Departing airport"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dapfeat_dict = make_onehot_feat_dict(flights_df, 'ORIGIN_AIRPORT', 'dap') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1e. Airline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "alfeat_dict = make_onehot_feat_dict(flights_df, 'AIRLINE', 'al') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1f. Departure hour"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Add departure hour as a feature\n",
    "flights_df['HOUR_DEPARTURE'] = np.floor(flights_df['SCHEDULED_DEPARTURE'].values/100).astype(int)\n",
    "hrfeat_dict = make_onehot_feat_dict(flights_df, 'HOUR_DEPARTURE', 'hr')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1g. Save new feature matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "all_dicts = [original_feat, daysfeat_dict, monthsfeat_dict, alfeat_dict, hrfeat_dict]\n",
    "feat_dict = all_dicts[0].copy()\n",
    "for d in all_dicts[1:]:\n",
    "    feat_dict.update(d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>f_al_AA</th>\n",
       "      <th>f_al_AS</th>\n",
       "      <th>f_al_B6</th>\n",
       "      <th>f_al_DL</th>\n",
       "      <th>f_al_EV</th>\n",
       "      <th>f_al_F9</th>\n",
       "      <th>f_al_HA</th>\n",
       "      <th>f_al_MQ</th>\n",
       "      <th>f_al_NK</th>\n",
       "      <th>f_al_OO</th>\n",
       "      <th>...</th>\n",
       "      <th>f_month_10</th>\n",
       "      <th>f_month_11</th>\n",
       "      <th>f_month_2</th>\n",
       "      <th>f_month_3</th>\n",
       "      <th>f_month_4</th>\n",
       "      <th>f_month_5</th>\n",
       "      <th>f_month_6</th>\n",
       "      <th>f_month_7</th>\n",
       "      <th>f_month_8</th>\n",
       "      <th>f_month_9</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 64 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  f_al_AA f_al_AS f_al_B6 f_al_DL f_al_EV f_al_F9 f_al_HA f_al_MQ f_al_NK  \\\n",
       "0   False   False   False   False   False   False   False   False   False   \n",
       "1   False    True   False   False   False   False   False   False   False   \n",
       "2   False   False   False   False   False   False   False   False   False   \n",
       "3   False   False   False   False   False   False   False   False   False   \n",
       "4   False   False   False   False    True   False   False   False   False   \n",
       "\n",
       "  f_al_OO    ...    f_month_10 f_month_11 f_month_2 f_month_3 f_month_4  \\\n",
       "0   False    ...         False      False     False     False     False   \n",
       "1   False    ...         False      False     False     False     False   \n",
       "2   False    ...         False      False     False     False     False   \n",
       "3   False    ...         False      False     False     False     False   \n",
       "4   False    ...         False      False     False     False     False   \n",
       "\n",
       "  f_month_5 f_month_6 f_month_7 f_month_8 f_month_9  \n",
       "0     False     False     False      True     False  \n",
       "1     False     False     False     False     False  \n",
       "2     False     False     False     False     False  \n",
       "3      True     False     False     False     False  \n",
       "4     False      True     False     False     False  \n",
       "\n",
       "[5 rows x 64 columns]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_feat = pd.DataFrame.from_dict(feat_dict)\n",
    "df_feat.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1f. Split features into training and validation sets for each airport"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "airport_keys = np.array(list(dapfeat_dict.keys()))\n",
    "X_train = {}\n",
    "X_test = {}\n",
    "y_train = {}\n",
    "y_test = {}\n",
    "for apk in airport_keys:\n",
    "    # Isolate X and y for each airport\n",
    "    ap_idx = np.transpose(np.argwhere(dapfeat_dict[apk]))[0]\n",
    "    X_all = df_feat.loc[ap_idx].values\n",
    "    y_all = labels_shuffle[ap_idx]\n",
    "    \n",
    "    # Calculate train and test set size\n",
    "    N_flights = len(y_all)\n",
    "    N_train = int(N_flights*.7)\n",
    "    N_test = N_flights - N_train\n",
    "    \n",
    "    # Make train and test sets\n",
    "    X_train[apk[-3:]] = X_all[:N_train]\n",
    "    X_test[apk[-3:]] = X_all[N_train:]\n",
    "    y_train[apk[-3:]] = y_all[:N_train]\n",
    "    y_test[apk[-3:]] = y_all[N_train:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Train model for each airport"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ABQ\n",
      "ALB\n",
      "ANC\n",
      "ATL\n",
      "AUS\n",
      "BDL\n",
      "BHM\n",
      "BNA\n",
      "BOI\n",
      "BOS\n",
      "BTR\n",
      "BUF\n",
      "BUR\n",
      "BWI\n",
      "CHS\n",
      "CID\n",
      "CLE\n",
      "CLT\n",
      "CMH\n",
      "COS\n",
      "CVG\n",
      "DAL\n",
      "DAY\n",
      "DCA\n",
      "DEN\n",
      "DFW\n",
      "DSM\n",
      "DTW\n",
      "ELP\n",
      "EWR\n",
      "FAT\n",
      "FLL\n",
      "GEG\n",
      "GRR\n",
      "GSO\n",
      "HNL\n",
      "HOU\n",
      "HPN\n",
      "IAD\n",
      "IAH\n",
      "ICT\n",
      "IND\n",
      "JAN\n",
      "JAX\n",
      "JFK\n",
      "KOA\n",
      "LAS\n",
      "LAX\n",
      "LGA\n",
      "LGB\n",
      "LIH\n",
      "LIT\n",
      "MAF\n",
      "MCI\n",
      "MCO\n",
      "MDW\n",
      "MEM\n",
      "MIA\n",
      "MKE\n",
      "MSN\n",
      "MSP\n",
      "MSY\n",
      "OAK\n",
      "OGG\n",
      "OKC\n",
      "OMA\n",
      "ONT\n",
      "ORD\n",
      "ORF\n",
      "PBI\n",
      "PDX\n",
      "PHL\n",
      "PHX\n",
      "PIT\n",
      "PSP\n",
      "PVD\n",
      "RDU\n",
      "RIC\n",
      "RNO\n",
      "ROC\n",
      "RSW\n",
      "SAN\n",
      "SAT\n",
      "SAV\n",
      "SDF\n",
      "SEA\n",
      "SFO\n",
      "SJC\n",
      "SJU\n",
      "SLC\n",
      "SMF\n",
      "SNA\n",
      "STL\n",
      "TPA\n",
      "TUL\n",
      "TUS\n",
      "TYS\n",
      "CPU times: user 42.1 s, sys: 2.3 s, total: 44.4 s\n",
      "Wall time: 16 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "train_aucs = {}\n",
    "test_aucs = {}\n",
    "train_ps = {}\n",
    "test_ps = {}\n",
    "C = 100000\n",
    "for k in X_train.keys():\n",
    "    print(k)\n",
    "    models = LogisticRegression(C=1)\n",
    "    models.fit(X_train[k], y_train[k])\n",
    "    \n",
    "    # Get probabilities\n",
    "    train_ps[k] = models.predict_proba(X_train[k])[:,1]\n",
    "    test_ps[k] = models.predict_proba(X_test[k])[:,1]\n",
    "\n",
    "    # Evaluate model\n",
    "    train_aucs[k] = roc_auc_score(y_train[k], train_ps[k])\n",
    "    test_aucs[k] = roc_auc_score(y_test[k], test_ps[k])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Evaluate predictor performance\n",
    "The area under the ROC curve (AUC) is used as the measure of accuracy."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3a. Training and validation accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC, validation set: 0.688660192785\n",
      "AUC, training set: 0.691104890377\n"
     ]
    }
   ],
   "source": [
    "all_ps = np.hstack(list(test_ps.values()))\n",
    "all_ys = np.hstack(list(y_test.values()))\n",
    "print('AUC, validation set:', roc_auc_score(all_ys, all_ps))\n",
    "\n",
    "all_ps_train = np.hstack(list(train_ps.values()))\n",
    "all_ys_train = np.hstack(list(y_train.values()))\n",
    "print('AUC, training set:', roc_auc_score(all_ys_train, all_ps_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3b. Plot distribution of accuracy across airports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x157f9dba8>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA/QAAAK6CAYAAACXABfKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nREREWogBPREREREREZEW+h89+apsgDCncwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1783cfe48>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 349,
       "width": 506
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot accuracy\n",
    "plt.figure(figsize=(8,5))\n",
    "accs = plt.hist(list(test_aucs.values()),bins=np.arange(0.6,0.8,0.02),color='k')\n",
    "plt.xlabel('accuracy (AUC)',size=20)\n",
    "plt.ylabel('count',size=20)\n",
    "plt.xticks(np.arange(0.6,0.8,0.02), size=15)\n",
    "plt.yticks(size=15)\n",
    "plt.title('Prediction accuracy across airports',size=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3c. Explore performance when adjust decision threshold\n",
    "Improve our ability to detect a flight that will be delayed while also predicting many false positives (predicting delayed when not delayed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix:\n",
      " [[1285267    3329]\n",
      " [ 279470    3868]]\n",
      "Precision:  0.537446158121\n",
      "Recall:  0.013651539857\n"
     ]
    }
   ],
   "source": [
    "decision_thresh = .5\n",
    "all_preds = all_ps > decision_thresh\n",
    "\n",
    "\n",
    "all_confuse = confusion_matrix(all_ys, all_preds)\n",
    "print('Confusion matrix:\\n',all_confuse)\n",
    "\n",
    "print('Precision: ', all_confuse[1,1]/(all_confuse[1,1]+all_confuse[0,1]))\n",
    "print('Recall: ', all_confuse[1,1]/(all_confuse[1,1]+all_confuse[1,0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix:\n",
      " [[1158143  130453]\n",
      " [ 206233   77105]]\n",
      "Precision:  0.371486524249\n",
      "Recall:  0.272130811963\n"
     ]
    }
   ],
   "source": [
    "decision_thresh = .3\n",
    "all_preds = all_ps > decision_thresh\n",
    "\n",
    "\n",
    "all_confuse = confusion_matrix(all_ys, all_preds)\n",
    "print('Confusion matrix:\\n',all_confuse)\n",
    "\n",
    "print('Precision: ', all_confuse[1,1]/(all_confuse[1,1]+all_confuse[0,1]))\n",
    "print('Recall: ', all_confuse[1,1]/(all_confuse[1,1]+all_confuse[1,0]))"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}