Henn-Notebook-FlyHomes-Kaggle-Zillow.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# FlyHomes Data Challenge: Kaggle Zillow Prize\n",
    "### Brian Henn - September 2018 \n",
    "\n",
    "### Key findings:\n",
    "\n",
    "-  A gradient boosted trees (GBT) approach is used to predict error in the Zillow estimate for home prices in Southern California\n",
    "- The model is based on a robust subset of the available features for each property that don't contain duplicate or missing data \n",
    "- The model is able to predict ~1% of the residual error in home prices in the Zestimate (for reference, Zillow Kaggle winners predicted about 8% of the residual error)\n",
    "- The most predictive features of Zestimate error are the tax assessed value of the home and its square footage \n",
    "- There is limited predictive signal of error (at least using this first-cut GBT approach) in these core features; further improvement might be gained by including noisier (more missing data) features or through more complex models\n",
    "\n",
    "### Notebook steps:\n",
    "\n",
    "1. Load data from sales files and extract properties with sales in 2016 and 2017\n",
    "2. Select features with robust, non-redundant data for inclusion in the model \n",
    "3. Test-train split on pooled 2016 and 2017 sales data\n",
    "4. Sample hyperparameter space of GBT model via cross validation on training data, select best one\n",
    "5. Train model on training data and then evaluate on test data\n",
    "6. Load overall property table and predict for all, write output file\n",
    "\n",
    "### Things to think about:\n",
    "\n",
    "- Need a method for dealing with time explicitly, as opposed to using all 2016-2017 data for predicting sales in that period: weight sales by recency in training? \n",
    "- Need a method to incorporate the rest of the features that are mostly missing data. Simply imputing the mean is not a good choice when 90+% of the column is missing. A KNN approach would likely be a good idea to impute values in these columns, then include the in the model, and produce better predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.preprocessing import Imputer\n",
    "from sklearn.model_selection import train_test_split, RandomizedSearchCV\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "from scipy.stats import randint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define filename and paths\n",
    "fname_features_2016 = './data/properties_2016.csv'\n",
    "fname_features_2017 = './data/properties_2017.csv'\n",
    "fname_sales_2016 = './data/train_2016_v2.csv'\n",
    "fname_sales_2017 = './data/train_2017.csv'\n",
    "fname_sub = './data/output_submission.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load 2016 and 2017 sales data\n",
    "sales_2016 = pd.read_csv(fname_sales_2016, index_col=0, header=0, \n",
    "                         parse_dates=[2], infer_datetime_format=True)\n",
    "sales_2017 = pd.read_csv(fname_sales_2017, index_col=0, header=0, \n",
    "                         parse_dates=[2], infer_datetime_format=True)\n",
    "sales = pd.concat([sales_2016, sales_2017])\n",
    "sales_set = sales.index.unique().values\n",
    "#print(sales.head(5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/brian/anaconda3/envs/insight/lib/python3.6/site-packages/IPython/core/interactiveshell.py:2903: DtypeWarning: Columns (49) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  if self.run_code(code, result):\n",
      "/home/brian/anaconda3/envs/insight/lib/python3.6/site-packages/IPython/core/interactiveshell.py:2903: DtypeWarning: Columns (32,34) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  if self.run_code(code, result):\n",
      "/home/brian/anaconda3/envs/insight/lib/python3.6/site-packages/IPython/core/interactiveshell.py:2903: DtypeWarning: Columns (22,49) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  if self.run_code(code, result):\n",
      "/home/brian/anaconda3/envs/insight/lib/python3.6/site-packages/IPython/core/interactiveshell.py:2903: DtypeWarning: Columns (22,32,34,49,55) have mixed types. Specify dtype option on import or set low_memory=False.\n",
      "  if self.run_code(code, result):\n"
     ]
    }
   ],
   "source": [
    "# inner join 2016 and 2017 feature data with sales data on parcel id\n",
    "\n",
    "# first, load 2016 feature data, grabbing only the rows with sales data for memory purposes\n",
    "iter_csv = pd.read_csv(fname_features_2016, index_col=0, header=0, iterator=True, chunksize=50000)\n",
    "features_2016 = pd.concat([chunk[chunk.index.isin(sales_set)] for chunk in iter_csv])\n",
    "\n",
    "# now, join the feature data to the sales data, duplicating properties with multiple sales\n",
    "features_2016_with_sales = features_2016.merge(sales, left_index=True, right_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fraction of rows with at least one missing value: 1.0000.\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Fraction of Missing Data')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x1080 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "frac_at_least_one_missing = sum(features_2016_with_sales.isnull().sum(1) > 0)/len(features_2016_with_sales)\n",
    "print('Fraction of rows with at least one missing value: %0.4f.\\n' % (frac_at_least_one_missing))\n",
    "\n",
    "frac_missing = []\n",
    "for col in features_2016_with_sales.columns.values:\n",
    "    frac_missing.append((len(features_2016_with_sales[col]) - features_2016_with_sales[col].count())/len(\n",
    "        features_2016_with_sales[col]))\n",
    "    \n",
    "fig = plt.figure()\n",
    "fig.set_size_inches([8,15])\n",
    "ax = fig.subplots()\n",
    "ax.barh(range(len(frac_missing)), frac_missing)\n",
    "ax.set_xlim([0,1]) \n",
    "ax.set_xlabel('Fraction Missing', fontsize=16)\n",
    "ax.set_ylim([0,len(frac_missing)])\n",
    "ax.set_xticks(np.arange(0,1.01,0.1))\n",
    "ax.set_yticks(range(len(frac_missing)))\n",
    "ax.set_yticklabels(features_2016_with_sales.columns.values,rotation = 0);\n",
    "ax.grid(True,'major','y')\n",
    "ax.set_title('Fraction of Missing Data', fontsize=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Takeaways from initial look at the data\n",
    "\n",
    "- Features seem to fall into two classes: robust (non-missing) and those that are mostly missing\n",
    "- Some features largely seem to duplicate each other (e.g., calculatedbathnbr vs. bathroomcnt) and so we can likely choose one\n",
    "- Some categorical features are going to be critical (property type) and we need a way of dealing with them\n",
    "\n",
    "So, let's subset to just features that do not have a lot of missing data, deal with the categoricals, and eliminate the redundant features, and go from there."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reduce features to those with at least 75% of data\n",
    "\n",
    "features_2016_with_sales = features_2016_with_sales.iloc[:,[frac < 0.25 for frac in frac_missing]]\n",
    "frac_missing = [frac_missing[i] for i, _ in enumerate(frac_missing) if frac_missing[i] < 0.25]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "261.0    112994\n",
      "266.0     42143\n",
      "246.0      4422\n",
      "269.0      4291\n",
      "248.0      1594\n",
      "247.0      1167\n",
      "265.0       691\n",
      "263.0       158\n",
      "275.0       105\n",
      "260.0        92\n",
      "267.0        58\n",
      "31.0         32\n",
      "264.0        16\n",
      "47.0          1\n",
      "Name: propertylandusetypeid, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "#print(features_2016_with_sales.shape)\n",
    "#print(features_2016_with_sales.columns)\n",
    "print(features_2016_with_sales['propertylandusetypeid'].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dealing with the categorical property type\n",
    "\n",
    "Since distinguishing between type (single-family, condo, etc.) should be an important feature of the model, let's break 'propertylandusetypeid' into categories based on the meaning of the code:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# transform categorical variable about property type into more useful set of dummy variables\n",
    "\n",
    "# single family homes (most of the dataset)\n",
    "features_2016_with_sales['single_family'] = [\n",
    "    1 if val == 261.0 else 0 for val in features_2016_with_sales['propertylandusetypeid']]\n",
    "# multi-family (duplex/triplex etc.)\n",
    "features_2016_with_sales['multi_family'] = [\n",
    "    1 if (val >= 246.0 and val <= 248.0) else 0 for val in features_2016_with_sales['propertylandusetypeid']]\n",
    "# condos (a lot of these also)\n",
    "features_2016_with_sales['condominium'] = [\n",
    "    1 if val == 266.0 else 0 for val in features_2016_with_sales['propertylandusetypeid']]\n",
    "# planned/cluster/mobile (a few of these)\n",
    "features_2016_with_sales['planned_community'] = [\n",
    "    1 if (val == 263.0 or val == 265.0 or val == 269.0) \\\n",
    "    else 0 for val in features_2016_with_sales['propertylandusetypeid']]\n",
    "# everything else (small number of random unusual residential types and commercial properties)\n",
    "features_2016_with_sales['other_property'] = [\n",
    "    1 if val not in [246.0, 247.0, 248.0, 261.0, 263.0, 265.0, 266.0, \n",
    "        269.0] else 0 for val in features_2016_with_sales['propertylandusetypeid']]\n",
    "\n",
    "#features_2016_with_sales[['propertylandusetypeid','single_family','multi_family','condominium']].sample(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Now, let's define the features we want in the model\n",
    "\n",
    "We've gotten rid of columns with mostly missing data, dealt with categoricals, and now we can look at the data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>bedroomcnt</th>\n",
       "      <th>bathroomcnt</th>\n",
       "      <th>calculatedfinishedsquarefeet</th>\n",
       "      <th>fullbathcnt</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>lotsizesquarefeet</th>\n",
       "      <th>single_family</th>\n",
       "      <th>multi_family</th>\n",
       "      <th>condominium</th>\n",
       "      <th>planned_community</th>\n",
       "      <th>other_property</th>\n",
       "      <th>regionidzip</th>\n",
       "      <th>yearbuilt</th>\n",
       "      <th>taxvaluedollarcnt</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>parcelid</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13842788</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1002.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>33736382.0</td>\n",
       "      <td>-117864598.0</td>\n",
       "      <td>7500.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>97001.0</td>\n",
       "      <td>1922.0</td>\n",
       "      <td>206647.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11933269</th>\n",
       "      <td>2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1226.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>34082100.0</td>\n",
       "      <td>-118282000.0</td>\n",
       "      <td>14999.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96007.0</td>\n",
       "      <td>1990.0</td>\n",
       "      <td>509000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12134938</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1922.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>34149866.0</td>\n",
       "      <td>-118029047.0</td>\n",
       "      <td>8281.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96267.0</td>\n",
       "      <td>1937.0</td>\n",
       "      <td>1010000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12865774</th>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1505.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>34109653.0</td>\n",
       "      <td>-117719071.0</td>\n",
       "      <td>5806.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96469.0</td>\n",
       "      <td>1953.0</td>\n",
       "      <td>436000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11344414</th>\n",
       "      <td>3.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1592.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>34473647.0</td>\n",
       "      <td>-118529858.0</td>\n",
       "      <td>4245.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96403.0</td>\n",
       "      <td>1989.0</td>\n",
       "      <td>297945.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17082998</th>\n",
       "      <td>3.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1457.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>34274183.0</td>\n",
       "      <td>-119244379.0</td>\n",
       "      <td>16007.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>97083.0</td>\n",
       "      <td>1948.0</td>\n",
       "      <td>77854.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12283597</th>\n",
       "      <td>3.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1345.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>33917029.0</td>\n",
       "      <td>-118194724.0</td>\n",
       "      <td>6297.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96113.0</td>\n",
       "      <td>1952.0</td>\n",
       "      <td>337000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11646043</th>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1155.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>34068000.0</td>\n",
       "      <td>-118453000.0</td>\n",
       "      <td>16349.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96005.0</td>\n",
       "      <td>1972.0</td>\n",
       "      <td>624268.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12420607</th>\n",
       "      <td>4.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2131.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>33852376.0</td>\n",
       "      <td>-118053568.0</td>\n",
       "      <td>4999.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96206.0</td>\n",
       "      <td>1972.0</td>\n",
       "      <td>379329.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13095617</th>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>809.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>34039825.0</td>\n",
       "      <td>-117918653.0</td>\n",
       "      <td>46098.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>96525.0</td>\n",
       "      <td>1973.0</td>\n",
       "      <td>238000.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          bedroomcnt  bathroomcnt  calculatedfinishedsquarefeet  fullbathcnt  \\\n",
       "parcelid                                                                       \n",
       "13842788         2.0          1.0                        1002.0          1.0   \n",
       "11933269         2.0          3.0                        1226.0          3.0   \n",
       "12134938         2.0          1.0                        1922.0          1.0   \n",
       "12865774         2.0          2.0                        1505.0          2.0   \n",
       "11344414         3.0          3.0                        1592.0          3.0   \n",
       "17082998         3.0          2.0                        1457.0          2.0   \n",
       "12283597         3.0          2.0                        1345.0          2.0   \n",
       "11646043         2.0          2.0                        1155.0          2.0   \n",
       "12420607         4.0          3.0                        2131.0          3.0   \n",
       "13095617         2.0          1.0                         809.0          1.0   \n",
       "\n",
       "            latitude    longitude  lotsizesquarefeet  single_family  \\\n",
       "parcelid                                                              \n",
       "13842788  33736382.0 -117864598.0             7500.0              1   \n",
       "11933269  34082100.0 -118282000.0            14999.0              0   \n",
       "12134938  34149866.0 -118029047.0             8281.0              1   \n",
       "12865774  34109653.0 -117719071.0             5806.0              1   \n",
       "11344414  34473647.0 -118529858.0             4245.0              1   \n",
       "17082998  34274183.0 -119244379.0            16007.0              1   \n",
       "12283597  33917029.0 -118194724.0             6297.0              1   \n",
       "11646043  34068000.0 -118453000.0            16349.0              0   \n",
       "12420607  33852376.0 -118053568.0             4999.0              1   \n",
       "13095617  34039825.0 -117918653.0            46098.0              0   \n",
       "\n",
       "          multi_family  condominium  planned_community  other_property  \\\n",
       "parcelid                                                                 \n",
       "13842788             0            0                  0               0   \n",
       "11933269             0            1                  0               0   \n",
       "12134938             0            0                  0               0   \n",
       "12865774             0            0                  0               0   \n",
       "11344414             0            0                  0               0   \n",
       "17082998             0            0                  0               0   \n",
       "12283597             0            0                  0               0   \n",
       "11646043             0            1                  0               0   \n",
       "12420607             0            0                  0               0   \n",
       "13095617             0            1                  0               0   \n",
       "\n",
       "          regionidzip  yearbuilt  taxvaluedollarcnt  \n",
       "parcelid                                             \n",
       "13842788      97001.0     1922.0           206647.0  \n",
       "11933269      96007.0     1990.0           509000.0  \n",
       "12134938      96267.0     1937.0          1010000.0  \n",
       "12865774      96469.0     1953.0           436000.0  \n",
       "11344414      96403.0     1989.0           297945.0  \n",
       "17082998      97083.0     1948.0            77854.0  \n",
       "12283597      96113.0     1952.0           337000.0  \n",
       "11646043      96005.0     1972.0           624268.0  \n",
       "12420607      96206.0     1972.0           379329.0  \n",
       "13095617      96525.0     1973.0           238000.0  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# list useful features for inclusion in the model\n",
    "\n",
    "features_in_model = ['bedroomcnt','bathroomcnt','calculatedfinishedsquarefeet','fullbathcnt',\n",
    "                     'latitude','longitude','lotsizesquarefeet','single_family','multi_family','condominium',\n",
    "                    'planned_community','other_property','regionidzip','yearbuilt','taxvaluedollarcnt']\n",
    "\n",
    "X = features_2016_with_sales[features_in_model]\n",
    "X.sample(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Impute missing data \n",
    "\n",
    "Since there are still missing values in these columns (<1% each), we can use a simple method (column mean) to fill them in.\n",
    "\n",
    "Now, we have a training and test dataset ready for modeling."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# impute missing values using sklearn's implementation\n",
    "\n",
    "# impute using the mean as a simple solution\n",
    "imputer = Imputer(strategy='mean')\n",
    "X_imputed = X.copy()\n",
    "for col in X.columns.values:\n",
    "    if sum(X[col].isna()) > 0: # only apply this to columns with missing data\n",
    "        imputer.fit(X[[col]])\n",
    "        X_imputed[col] = imputer.transform(X[[col]]).ravel()\n",
    "\n",
    "X_imputed.sample(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get sales data as training dataset\n",
    "\n",
    "Y = features_2016_with_sales[['logerror','transactiondate']]\n",
    "#Y.sample(15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Examine feature correlations and correlations with target\n",
    "\n",
    "We want to check for correlations bewteen our features and between the features and the target (Zestimate log error). If features are strongly colinear we maybe should not include them. And the correlations between the features and the target may give a sense of what will predict what we care about.\n",
    "\n",
    "To do this we will make a heat map of correlations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fe19398aba8>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x648 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# examine the correlations among the selected features and with the log error\n",
    "\n",
    "# https://seaborn.pydata.org/examples/many_pairwise_correlations.html\n",
    "\n",
    "# set up colormap for correlation plots\n",
    "# Generate a custom diverging colormap\n",
    "cmap = sns.diverging_palette(220, 10, as_cmap=True)\n",
    "cmax = 1\n",
    "cmin = -1\n",
    "\n",
    "# create a dataframe of just the features we want to look at, plus the target)\n",
    "plot_columns = features_in_model\n",
    "plot_columns.append('logerror')\n",
    "full_df = features_2016_with_sales[plot_columns]\n",
    "                                                    \n",
    "# Compute the correlation matrix\n",
    "corr_properties = full_df.corr()\n",
    "\n",
    "# Generate a mask for the upper triangle\n",
    "mask = np.zeros_like(corr_properties, dtype=np.bool)\n",
    "mask[np.triu_indices_from(mask)] = True\n",
    "\n",
    "# Set up the matplotlib figure\n",
    "f, ax = plt.subplots(figsize=(11, 9))\n",
    "\n",
    "# Draw the heatmap with the mask and correct aspect ratio\n",
    "sns.heatmap(corr_properties, mask=mask, cmap=cmap, vmax=cmax, vmin=cmin, center=0,\n",
    "            square=True, linewidths=.5, cbar_kws={\"shrink\": .5})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(16, 16)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# now bar plot just correlation of logerror against features\n",
    "\n",
    "print(corr_properties.shape)\n",
    "\n",
    "fig = plt.figure()\n",
    "fig.set_size_inches([8,6])\n",
    "ax = fig.subplots()\n",
    "ax.bar(range(len(full_df.columns.values) - 1), corr_properties.iloc[-1,0:15])\n",
    "ax.set_ylim([-0.25,0.25])\n",
    "ax.set_ylabel('Correlation with Log Error', fontsize=16)\n",
    "ax.set_xticks(range(len(full_df.columns.values) - 1))\n",
    "ax.set_xticklabels(full_df.columns.values[0:-1],rotation=-90, fontsize=16)\n",
    "ax.grid(True,'major','y')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Takeaways from correlation analysis\n",
    "\n",
    "- The features are not strongly co-linear, which is good. The only ones that are co-linear are things like bathroom count vs. full bathroom count, or between the property categories, which is to be expected. We still might expect added information from these categories so we keep them.\n",
    "- There is little correlation between the features and the log error. This is not surprising since if there were then making the Zestimate better wold be an easy problem to solve! However, this is linear correlation; it is possible that there is structure in the log error that a non-linear model like GBT can pick up on.\n",
    "\n",
    "So, let's proceed with model training on these features!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split sample into training and test sets \n",
    "# retain 20% of the properties with sales (34k) as a test dataset, train on 80% (134k)\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X_imputed, Y['logerror'], test_size=0.2, shuffle=True)\n",
    "\n",
    "# Note: for the moment we are not worry about the time dimension of the sales,\n",
    "# instead just using all of the sales from 2016 and 2017 as both training and validation data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Choosing the right model: Gradient Boosting Regressor\n",
    "\n",
    "I choose a GBT model for this problem because:\n",
    "\n",
    "- it can handle nonlinear relationships between features and targets\n",
    "- it can avoiding overfitting using an stochastic ensemble approach of boosted trees\n",
    "- it has been shown to have high skill for many ML problems \n",
    "- it can produce confidence intervals on predictions directly from the ensemble approach\n",
    "\n",
    "To make the GBT model work well, I apply the following approach:\n",
    "\n",
    "- tuning of hyperparameters via k-fold cross validation against the training dataset\n",
    "- including multiple hyperparameters (number of boosted trees, depth of trees, split metric, etc.) in the tuning\n",
    "- estimating error on the test dataset using a model trained using the best hyperparameters on the training dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 20 candidates, totalling 100 fits\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97 \n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97 \n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97 \n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97, score=0.0020284824094071174, total= 1.5min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97, score=0.0024106475403411487, total= 1.6min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97, score=-0.001270045211486881, total= 1.6min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97, score=0.0009893498179117666, total= 1.6min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67, score=0.0004961601270079274, total=  51.7s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67, score=0.0011869594820169693, total=  52.5s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67, score=-0.0007592732690144821, total=  52.9s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=1, min_samples_split=6, n_estimators=97, score=0.003019989729796002, total= 1.7min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67, score=0.0016367650104135434, total=  50.9s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=4, n_estimators=67, score=0.002223608853322956, total=  50.3s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79, score=0.0006220446789639844, total= 1.0min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79, score=0.0012468608702934647, total=  58.3s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2, score=-0.0019293590643856273, total=   1.3s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2, score=-0.002296001877972742, total=   1.3s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2, score=-0.002725307082784179, total=   1.1s\n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79, score=-0.0007539189411271964, total=  59.0s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2 \n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79, score=0.0018965663888143025, total=  59.0s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35 \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  17 tasks      | elapsed:  4.3min\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2, score=-0.0006972801266680317, total=   1.5s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=4, min_samples_split=6, n_estimators=2, score=-0.0015648071857161663, total=   1.4s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35, score=-0.0004596602601187882, total=  12.9s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35, score=8.891360863128028e-05, total=  12.3s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35, score=-0.001271771519731013, total=  13.9s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=5, min_samples_split=6, n_estimators=79, score=0.002370937641277382, total= 1.1min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35, score=0.0011019909719853782, total=  11.4s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=3, n_estimators=35, score=0.001204061791511002, total=  12.1s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81, score=0.0017289003595403196, total= 1.5min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81, score=0.0007620158248112796, total= 1.6min\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81, score=0.0018735523942794428, total= 1.6min\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81, score=-0.0009786146723402833, total= 1.6min\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60, score=0.00036929545651520623, total=  48.5s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60, score=0.0009442884275997488, total=  49.7s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60, score=-0.0013580307579112993, total=  51.0s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=81, score=0.0028460418117346453, total= 1.5min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60, score=0.0016571509522248506, total=  47.8s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=7, n_estimators=60, score=0.00216606076846082, total=  48.0s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89, score=0.0004723215734770969, total= 1.2min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89, score=0.0014327786988900248, total= 1.2min\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89, score=-0.0008065770566014763, total= 1.2min\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7, score=-0.0008742364154183946, total=   8.5s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7, score=-0.0016836271661344606, total=   8.6s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7, score=-0.0011405370258570358, total=   9.5s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89, score=0.0017776984873241197, total= 1.2min\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7, score=-6.200289748536925e-05, total=   9.3s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=6, n_estimators=7, score=0.00047069484339756773, total=  10.3s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84, score=0.00037981891505700643, total=  27.6s\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=2, min_samples_split=9, n_estimators=89, score=0.0024536326590187674, total= 1.3min\n",
      "[CV] criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84, score=0.0017335809859008176, total=  28.0s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84, score=-0.0005278948439475517, total=  29.6s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84, score=0.0018808235703613452, total=  25.8s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=sqrt, min_samples_leaf=7, min_samples_split=10, n_estimators=84, score=0.0019483601759190394, total=  27.9s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92, score=0.0021855149579713684, total= 1.7min\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92, score=0.0010264446797636495, total= 1.7min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92, score=0.00016397533847234416, total= 1.7min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92, score=0.002164844906696217, total= 1.7min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73, score=0.000626843091413809, total=  27.9s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73, score=0.001944227096265072, total=  29.3s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73, score=0.00043724236257070004, total=  28.7s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11, score=-0.0008219691383106743, total=   5.2s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73, score=0.0020237789015241647, total=  29.9s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11, score=-0.0004882792714198114, total=   4.8s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11, score=-0.0015486663171253578, total=   5.5s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11, score=0.00038548561378415336, total=   4.8s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=5, n_estimators=11, score=0.0001290463431099509, total=   5.4s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=6, min_samples_split=4, n_estimators=73, score=0.002489510895604341, total=  30.1s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86, score=0.0003382504585619506, total=  36.4s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86, score=0.0054362843734396815, total=  35.3s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86, score=-0.00030646155718438983, total=  36.3s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=3, n_estimators=92, score=0.0033671480860546503, total= 2.0min\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45, score=-0.00036182183148136815, total=  15.6s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45, score=0.0009289597001088666, total=  16.6s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86, score=0.0033028491342972366, total=  39.4s\n",
      "[CV] criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=7, n_estimators=86, score=0.0026301092187526187, total=  39.0s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45, score=-0.0012294019841627701, total=  16.2s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45, score=0.0012988217933675994, total=  15.8s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23 \n",
      "[CV]  criterion=friedman_mse, max_depth=3, max_features=sqrt, min_samples_leaf=5, min_samples_split=3, n_estimators=45, score=0.0015580715194315475, total=  14.5s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23, score=-0.00024297426928354682, total=  20.6s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23, score=5.8000782924061405e-05, total=  19.8s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23, score=-0.001271646519723335, total=  19.6s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23, score=0.0011172512985517802, total=  21.2s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=6, min_samples_split=3, n_estimators=23, score=0.0010231622067909951, total=  21.0s\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55, score=0.0010018766947678115, total= 1.1min\n",
      "[CV] criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55, score=0.0017581171650148386, total= 1.1min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55, score=-0.0002666316474033703, total= 1.1min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55, score=0.0018366879200171216, total= 1.1min\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49, score=0.0005596359408218676, total=  18.0s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49, score=0.001811350142062129, total=  19.3s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49, score=-0.00029998379886664317, total=  21.0s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49, score=0.0017393400511360735, total=  20.9s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21, score=-0.0003828955987605909, total=  10.2s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21, score=-4.900257029927246e-06, total=  10.0s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21 \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  90 tasks      | elapsed: 16.4min\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=3, min_samples_split=4, n_estimators=49, score=0.002293414689347073, total=  22.7s\n",
      "[CV] criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21, score=-0.0011538296894566535, total=   9.2s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21, score=0.0010570687102220422, total=   9.7s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56 \n",
      "[CV]  criterion=friedman_mse, max_depth=4, max_features=sqrt, min_samples_leaf=7, min_samples_split=7, n_estimators=21, score=0.0008502412052023844, total=   9.9s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56 \n",
      "[CV]  criterion=mse, max_depth=4, max_features=None, min_samples_leaf=10, min_samples_split=2, n_estimators=55, score=0.002665783165695213, total= 1.2min\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56, score=0.0004682828716400244, total=  46.3s\n",
      "[CV] criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56 \n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56, score=0.000985206995143706, total=  46.6s\n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56, score=-0.0009059397771797872, total=  45.7s\n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56, score=0.0016236381290689605, total=  43.5s\n",
      "[CV]  criterion=mse, max_depth=3, max_features=None, min_samples_leaf=7, min_samples_split=8, n_estimators=56, score=0.0020151803586979655, total=  20.3s\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done 100 out of 100 | elapsed: 17.7min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise',\n",
       "          estimator=GradientBoostingRegressor(alpha=0.9, criterion='friedman_mse', init=None,\n",
       "             learning_rate=0.1, loss='lad', max_depth=3, max_features=None,\n",
       "             max_leaf_nodes=None, min_impurity_decrease=0.0,\n",
       "             min_impurity_split=None, min_samples_leaf=1,\n",
       "             min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "             n_estimators=100, presort='auto', random_state=None,\n",
       "             subsample=1.0, verbose=0, warm_start=False),\n",
       "          fit_params=None, iid=True, n_iter=20, n_jobs=-1,\n",
       "          param_distributions={'n_estimators': <scipy.stats._distn_infrastructure.rv_frozen object at 0x7fe18e9b2358>, 'max_depth': <scipy.stats._distn_infrastructure.rv_frozen object at 0x7fe18e9b24a8>, 'max_features': [None, 'sqrt'], 'min_samples_split': <scipy.stats._distn_infrastructure.rv_frozen object at 0x7fe18e9b2630>, 'min_samples_leaf': <scipy.stats._distn_infrastructure.rv_frozen object at 0x7fe18e9b2908>, 'criterion': ['friedman_mse', 'mse']},\n",
       "          pre_dispatch='2*n_jobs', random_state=120914, refit=True,\n",
       "          return_train_score=True, scoring=None, verbose=4)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's train the gradient boosted tree model's hyperparameters using sklearn's RandomSearchCV capabilities\n",
    "\n",
    "# define the estimator as GBT\n",
    "gbt = GradientBoostingRegressor(loss='lad')\n",
    "\n",
    "# define the hyperparameter we want to search across in random sampling (as in section 2)\n",
    "param_dist = {'n_estimators' : randint(2, 100), # number of boosting iterations \n",
    "              'max_depth' : randint(3, 5), # allow for potentially many (100 splits) in the trees\n",
    "              'max_features' : [None, 'sqrt'], # allow either any number of features in bagging, or just the sqrt\n",
    "              'min_samples_split' : randint(2, 11), # controls on splitting and leaves\n",
    "              'min_samples_leaf' : randint(1, 11),\n",
    "              'criterion' : ['friedman_mse', 'mse']} # vary the scoring method\n",
    "# random sampling strategy suggested by scikit-learn.org/stable/auto_examples/model_selection/plot_randomized_search.html#sphx-glr-auto-examples-model-selection-plot-randomized-search-py\n",
    "\n",
    "# randomly search parameters using my laptop's cores\n",
    "n_iter_random = 20 # 20 iterations of parameters for each fold \n",
    "cv_random = 5 # 5-fold cross validation\n",
    "gbt_hypertuning = RandomizedSearchCV(estimator=gbt, param_distributions=param_dist, n_iter=n_iter_random,\n",
    "                               cv=cv_random, random_state=120914, n_jobs=-1, verbose=4, return_train_score=True)\n",
    "\n",
    "# Fit the random search model for hyperparameter tuning\n",
    "gbt_hypertuning.fit(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_criterion</th>\n",
       "      <th>param_max_depth</th>\n",
       "      <th>param_max_features</th>\n",
       "      <th>param_min_samples_leaf</th>\n",
       "      <th>param_min_samples_split</th>\n",
       "      <th>param_n_estimators</th>\n",
       "      <th>...</th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>rank_test_score</th>\n",
       "      <th>split0_train_score</th>\n",
       "      <th>split1_train_score</th>\n",
       "      <th>split2_train_score</th>\n",
       "      <th>split3_train_score</th>\n",
       "      <th>split4_train_score</th>\n",
       "      <th>mean_train_score</th>\n",
       "      <th>std_train_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>36.948208</td>\n",
       "      <td>1.554087</td>\n",
       "      <td>0.319071</td>\n",
       "      <td>0.086717</td>\n",
       "      <td>friedman_mse</td>\n",
       "      <td>4</td>\n",
       "      <td>sqrt</td>\n",
       "      <td>3</td>\n",
       "      <td>7</td>\n",
       "      <td>86</td>\n",
       "      <td>...</td>\n",
       "      <td>0.002280</td>\n",
       "      <td>0.002078</td>\n",
       "      <td>1</td>\n",
       "      <td>0.005522</td>\n",
       "      <td>0.004334</td>\n",
       "      <td>0.004557</td>\n",
       "      <td>0.005122</td>\n",
       "      <td>0.003976</td>\n",
       "      <td>0.004702</td>\n",
       "      <td>0.000553</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>105.957788</td>\n",
       "      <td>6.320785</td>\n",
       "      <td>0.253485</td>\n",
       "      <td>0.062552</td>\n",
       "      <td>mse</td>\n",
       "      <td>4</td>\n",
       "      <td>None</td>\n",
       "      <td>10</td>\n",
       "      <td>3</td>\n",
       "      <td>92</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001782</td>\n",
       "      <td>0.001096</td>\n",
       "      <td>2</td>\n",
       "      <td>0.004830</td>\n",
       "      <td>0.004393</td>\n",
       "      <td>0.005439</td>\n",
       "      <td>0.004858</td>\n",
       "      <td>0.004324</td>\n",
       "      <td>0.004769</td>\n",
       "      <td>0.000400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>28.926908</td>\n",
       "      <td>0.803978</td>\n",
       "      <td>0.242315</td>\n",
       "      <td>0.051427</td>\n",
       "      <td>friedman_mse</td>\n",
       "      <td>4</td>\n",
       "      <td>sqrt</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>73</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001504</td>\n",
       "      <td>0.000818</td>\n",
       "      <td>3</td>\n",
       "      <td>0.004813</td>\n",
       "      <td>0.003312</td>\n",
       "      <td>0.004833</td>\n",
       "      <td>0.003336</td>\n",
       "      <td>0.002941</td>\n",
       "      <td>0.003847</td>\n",
       "      <td>0.000809</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>95.640701</td>\n",
       "      <td>3.652272</td>\n",
       "      <td>0.187319</td>\n",
       "      <td>0.035003</td>\n",
       "      <td>friedman_mse</td>\n",
       "      <td>4</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "      <td>6</td>\n",
       "      <td>97</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001436</td>\n",
       "      <td>0.001505</td>\n",
       "      <td>4</td>\n",
       "      <td>0.005946</td>\n",
       "      <td>0.004634</td>\n",
       "      <td>0.006714</td>\n",
       "      <td>0.006619</td>\n",
       "      <td>0.005238</td>\n",
       "      <td>0.005830</td>\n",
       "      <td>0.000800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>66.315540</td>\n",
       "      <td>2.036633</td>\n",
       "      <td>0.130601</td>\n",
       "      <td>0.018190</td>\n",
       "      <td>mse</td>\n",
       "      <td>4</td>\n",
       "      <td>None</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "      <td>55</td>\n",
       "      <td>...</td>\n",
       "      <td>0.001399</td>\n",
       "      <td>0.000986</td>\n",
       "      <td>5</td>\n",
       "      <td>0.003873</td>\n",
       "      <td>0.003270</td>\n",
       "      <td>0.003837</td>\n",
       "      <td>0.003094</td>\n",
       "      <td>0.003119</td>\n",
       "      <td>0.003439</td>\n",
       "      <td>0.000346</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 26 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "13      36.948208      1.554087         0.319071        0.086717   \n",
       "10     105.957788      6.320785         0.253485        0.062552   \n",
       "11      28.926908      0.803978         0.242315        0.051427   \n",
       "0       95.640701      3.652272         0.187319        0.035003   \n",
       "16      66.315540      2.036633         0.130601        0.018190   \n",
       "\n",
       "   param_criterion param_max_depth param_max_features param_min_samples_leaf  \\\n",
       "13    friedman_mse               4               sqrt                      3   \n",
       "10             mse               4               None                     10   \n",
       "11    friedman_mse               4               sqrt                      6   \n",
       "0     friedman_mse               4               None                      1   \n",
       "16             mse               4               None                     10   \n",
       "\n",
       "   param_min_samples_split param_n_estimators       ...         \\\n",
       "13                       7                 86       ...          \n",
       "10                       3                 92       ...          \n",
       "11                       4                 73       ...          \n",
       "0                        6                 97       ...          \n",
       "16                       2                 55       ...          \n",
       "\n",
       "   mean_test_score  std_test_score  rank_test_score  split0_train_score  \\\n",
       "13        0.002280        0.002078                1            0.005522   \n",
       "10        0.001782        0.001096                2            0.004830   \n",
       "11        0.001504        0.000818                3            0.004813   \n",
       "0         0.001436        0.001505                4            0.005946   \n",
       "16        0.001399        0.000986                5            0.003873   \n",
       "\n",
       "    split1_train_score  split2_train_score  split3_train_score  \\\n",
       "13            0.004334            0.004557            0.005122   \n",
       "10            0.004393            0.005439            0.004858   \n",
       "11            0.003312            0.004833            0.003336   \n",
       "0             0.004634            0.006714            0.006619   \n",
       "16            0.003270            0.003837            0.003094   \n",
       "\n",
       "    split4_train_score  mean_train_score  std_train_score  \n",
       "13            0.003976          0.004702         0.000553  \n",
       "10            0.004324          0.004769         0.000400  \n",
       "11            0.002941          0.003847         0.000809  \n",
       "0             0.005238          0.005830         0.000800  \n",
       "16            0.003119          0.003439         0.000346  \n",
       "\n",
       "[5 rows x 26 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_df = pd.DataFrame(gbt_hypertuning.cv_results_)\n",
    "results_df.sort_values('mean_test_score', ascending=False, inplace=True)\n",
    "results_df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'criterion': 'friedman_mse', 'max_depth': 4, 'max_features': 'sqrt', 'min_samples_leaf': 3, 'min_samples_split': 7, 'n_estimators': 86}\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GradientBoostingRegressor(alpha=0.9, criterion='friedman_mse', init=None,\n",
       "             learning_rate=0.1, loss='ls', max_depth=4,\n",
       "             max_features='sqrt', max_leaf_nodes=None,\n",
       "             min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "             min_samples_leaf=3, min_samples_split=7,\n",
       "             min_weight_fraction_leaf=0.0, n_estimators=86, presort='auto',\n",
       "             random_state=None, subsample=1.0, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# train model using best hyperparameters and whole training dataset\n",
    "\n",
    "# use the best hyperparameters\n",
    "best = gbt_hypertuning.best_params_\n",
    "print(best)\n",
    "gbt_train = GradientBoostingRegressor(**best)\n",
    "\n",
    "# train the model\n",
    "gbt_train.fit(X_train,Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.06972498957911655\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.014051613881057912"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# evaluate best model on test dataset\n",
    "\n",
    "# produce predictions \n",
    "Y_pred = gbt_train.predict(X_test)\n",
    "\n",
    "# compute MAE on predictions\n",
    "MAE = mean_absolute_error(Y_pred, Y_test)\n",
    "print(MAE)\n",
    "\n",
    "gbt_train.score(X_test, Y_test, sample_weight=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Model Evaluation Plot')"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots()\n",
    "fig.set_size_inches([12,8])\n",
    "#ax.plot(Y_test, Y_pred, 'b +')\n",
    "ax = sns.kdeplot(Y_test[::10], Y_pred[::10], shade=True)\n",
    "ax.set_xlim([-0.1, 0.1])\n",
    "ax.set_ylim([-0.1, 0.1])\n",
    "ax.set_xlabel('Actual Log Error', fontsize=16)\n",
    "ax.set_ylabel('Predicted Log Error', fontsize=16)\n",
    "ax.grid()\n",
    "ax.set_title('Model Evaluation Plot')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpretation of model training\n",
    "\n",
    "Well, that's not great, but it's something. The tuned/trained model was able to predict about 1.4% of the variance (R^2 = 0.014) of the log error. In other words, of the residual error in the Zestimate, this model can predict about 1/70 of it. So if the current Zestimate has an error of about +/- 7% as implied by the mean absolute log error of 0.069, then an improvement such as this would reduce that by about 0.10 percentage points, say to +/- 6.9%. Better than nothing I guess!\n",
    "\n",
    "For reference, the top of the kaggle leaderboard shows models that reduce the log error AME to about 0.064, indicating that they were able to predict perhaps 8% of the variance of the residual error. That would be more like an improvement in Zestimate error to +/- 6.4%. So while those models are better than what was produced here, they are not dramatically better. **This is a hard problem!**\n",
    "\n",
    "Let's now examine the feature importance in the model to see what actually predicted Zestimate error."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,0,'Feature Name')"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# now extract feature importances and plot\n",
    "feature_importances = gbt_train.feature_importances_\n",
    "feature_names = X_imputed.columns.values\n",
    "\n",
    "fig = plt.figure()\n",
    "fig.set_size_inches([12,4])\n",
    "ax = fig.subplots(1)\n",
    "ax.bar(range(1,len(feature_importances) + 1), 100*feature_importances/max(feature_importances))\n",
    "ax.set_ylim([0, 100])\n",
    "ax.set_xticks(range(1,len(feature_importances) + 1))\n",
    "ax.set_xticklabels(feature_names,FontSize=16, rotation=90)\n",
    "ax.grid(True,'major','y')\n",
    "ax.set_ylabel('Feature Importance',FontSize=16)\n",
    "ax.set_xlabel('Feature Name',FontSize=16)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature importance interpretation\n",
    "\n",
    "The two most important features are the tax payment and the square footage. Location, year built, and lot size show some importance also. The type of property was not predictive (intersting).\n",
    "\n",
    "So, we can interpret that the Zestimate model (which uses price per square foot in its calculations I believe) is sensitive to errors due to this assumption. It also shows that the way that the assessed tax value/payment is used may introduce errors. Both of these features would therefore be useful starting places in improving the model.\n",
    "\n",
    "### A note about time \n",
    "\n",
    "In this very preliminary approach I did not account for time explicitly, instead using the entire 2016-2017 sales data as training for the features in the 2016 dataset. This muddles the effect of time and may limit the skill of the model, i.e., all 6 of the prediction periods have the same forecast in this model. \n",
    "\n",
    "A logical next step would be to explicity account for the timing of recent sales in making predictions for a given time. The way to do this might be to weight the sales data by its recency to the forecast date during model training, and excludes sales data from *after* the forecast date. This obviously would have to happen in a real model implementation. \n",
    "\n",
    "#### The rest of the notebook deals with predicting the full (3M) property dataset and producing an output file for submission to Kaggle. FWIW, this notebook produced a middle/back of the pack solution relative to the public leaderboard. An interesting problem I would like to spend more time on :)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/brian/anaconda3/envs/insight/lib/python3.6/site-packages/numpy/lib/arraysetops.py:472: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison\n",
      "  mask |= (ar1 == a)\n"
     ]
    }
   ],
   "source": [
    "# in order to make predictions for the full dataset, first load the model columns from the 2016 data\n",
    "\n",
    "raw_features_in_model = ['parcelid','bedroomcnt','bathroomcnt','calculatedfinishedsquarefeet','fullbathcnt',\n",
    "                         'latitude','longitude','lotsizesquarefeet','propertylandusetypeid',\n",
    "                         'regionidzip','yearbuilt','taxvaluedollarcnt']\n",
    "\n",
    "features_full_2016 = pd.read_csv(fname_features_2016, index_col=0, header=0, usecols=raw_features_in_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>bathroomcnt</th>\n",
       "      <th>bedroomcnt</th>\n",
       "      <th>calculatedfinishedsquarefeet</th>\n",
       "      <th>fullbathcnt</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>lotsizesquarefeet</th>\n",
       "      <th>propertylandusetypeid</th>\n",
       "      <th>regionidzip</th>\n",
       "      <th>yearbuilt</th>\n",
       "      <th>taxvaluedollarcnt</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>parcelid</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>13981101</th>\n",
       "      <td>3.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>3508.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>33633567.0</td>\n",
       "      <td>-117891221.0</td>\n",
       "      <td>10720.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96978.0</td>\n",
       "      <td>1965.0</td>\n",
       "      <td>2217000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11490231</th>\n",
       "      <td>3.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1830.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>33869700.0</td>\n",
       "      <td>-118379000.0</td>\n",
       "      <td>7493.0</td>\n",
       "      <td>266.0</td>\n",
       "      <td>96124.0</td>\n",
       "      <td>2004.0</td>\n",
       "      <td>746517.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13943549</th>\n",
       "      <td>2.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1749.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>33763110.0</td>\n",
       "      <td>-117980464.0</td>\n",
       "      <td>9112.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96990.0</td>\n",
       "      <td>1959.0</td>\n",
       "      <td>450828.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13113569</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1422.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>33984100.0</td>\n",
       "      <td>-118454000.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>266.0</td>\n",
       "      <td>96128.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>367845.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11347912</th>\n",
       "      <td>4.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>3389.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>34471635.0</td>\n",
       "      <td>-118554916.0</td>\n",
       "      <td>9408.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96377.0</td>\n",
       "      <td>2003.0</td>\n",
       "      <td>681000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12613076</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>33790208.0</td>\n",
       "      <td>-118313609.0</td>\n",
       "      <td>9179.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96217.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>33682.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11491300</th>\n",
       "      <td>4.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1452.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>33870261.0</td>\n",
       "      <td>-118368065.0</td>\n",
       "      <td>7507.0</td>\n",
       "      <td>266.0</td>\n",
       "      <td>96124.0</td>\n",
       "      <td>1999.0</td>\n",
       "      <td>637486.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13903176</th>\n",
       "      <td>1.5</td>\n",
       "      <td>3.0</td>\n",
       "      <td>1318.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>33879462.0</td>\n",
       "      <td>-117991706.0</td>\n",
       "      <td>1666.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96181.0</td>\n",
       "      <td>1973.0</td>\n",
       "      <td>311272.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11938228</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1424.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>34065942.0</td>\n",
       "      <td>-118237471.0</td>\n",
       "      <td>1350.0</td>\n",
       "      <td>47.0</td>\n",
       "      <td>95993.0</td>\n",
       "      <td>1947.0</td>\n",
       "      <td>68691.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10783987</th>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1016.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>34192696.0</td>\n",
       "      <td>-118549671.0</td>\n",
       "      <td>6242.0</td>\n",
       "      <td>261.0</td>\n",
       "      <td>96364.0</td>\n",
       "      <td>1951.0</td>\n",
       "      <td>366000.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          bathroomcnt  bedroomcnt  calculatedfinishedsquarefeet  fullbathcnt  \\\n",
       "parcelid                                                                       \n",
       "13981101          3.0         4.0                        3508.0          3.0   \n",
       "11490231          3.0         3.0                        1830.0          3.0   \n",
       "13943549          2.0         3.0                        1749.0          2.0   \n",
       "13113569          0.0         0.0                        1422.0          NaN   \n",
       "11347912          4.0         5.0                        3389.0          4.0   \n",
       "12613076          0.0         0.0                           NaN          NaN   \n",
       "11491300          4.0         3.0                        1452.0          4.0   \n",
       "13903176          1.5         3.0                        1318.0          1.0   \n",
       "11938228          0.0         0.0                        1424.0          NaN   \n",
       "10783987          1.0         2.0                        1016.0          1.0   \n",
       "\n",
       "            latitude    longitude  lotsizesquarefeet  propertylandusetypeid  \\\n",
       "parcelid                                                                      \n",
       "13981101  33633567.0 -117891221.0            10720.0                  261.0   \n",
       "11490231  33869700.0 -118379000.0             7493.0                  266.0   \n",
       "13943549  33763110.0 -117980464.0             9112.0                  261.0   \n",
       "13113569  33984100.0 -118454000.0                NaN                  266.0   \n",
       "11347912  34471635.0 -118554916.0             9408.0                  261.0   \n",
       "12613076  33790208.0 -118313609.0             9179.0                  261.0   \n",
       "11491300  33870261.0 -118368065.0             7507.0                  266.0   \n",
       "13903176  33879462.0 -117991706.0             1666.0                  261.0   \n",
       "11938228  34065942.0 -118237471.0             1350.0                   47.0   \n",
       "10783987  34192696.0 -118549671.0             6242.0                  261.0   \n",
       "\n",
       "          regionidzip  yearbuilt  taxvaluedollarcnt  \n",
       "parcelid                                             \n",
       "13981101      96978.0     1965.0          2217000.0  \n",
       "11490231      96124.0     2004.0           746517.0  \n",
       "13943549      96990.0     1959.0           450828.0  \n",
       "13113569      96128.0        NaN           367845.0  \n",
       "11347912      96377.0     2003.0           681000.0  \n",
       "12613076      96217.0        NaN            33682.0  \n",
       "11491300      96124.0     1999.0           637486.0  \n",
       "13903176      96181.0     1973.0           311272.0  \n",
       "11938228      95993.0     1947.0            68691.0  \n",
       "10783987      96364.0     1951.0           366000.0  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features_full_2016.sample(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# repeat dealing with categorical data and imputation \n",
    "\n",
    "# single family homes (most of the dataset)\n",
    "features_full_2016['single_family'] = [\n",
    "    1 if val == 261.0 else 0 for val in features_full_2016['propertylandusetypeid']]\n",
    "# multi-family (duplex/triplex etc.)\n",
    "features_full_2016['multi_family'] = [\n",
    "    1 if (val >= 246.0 and val <= 248.0) else 0 for val in features_full_2016['propertylandusetypeid']]\n",
    "# condos (a lot of these also)\n",
    "features_full_2016['condominium'] = [\n",
    "    1 if val == 266.0 else 0 for val in features_full_2016['propertylandusetypeid']]\n",
    "# planned/cluster/mobile (a few of these)\n",
    "features_full_2016['planned_community'] = [\n",
    "    1 if (val == 263.0 or val == 265.0 or val == 269.0) \\\n",
    "    else 0 for val in features_full_2016['propertylandusetypeid']]\n",
    "# everything else (small number of random unusual residential types and commercial properties)\n",
    "features_full_2016['other_property'] = [\n",
    "    1 if val not in [246.0, 247.0, 248.0, 261.0, 263.0, 265.0, 266.0, \n",
    "        269.0] else 0 for val in features_full_2016['propertylandusetypeid']]\n",
    "\n",
    "X_full = features_full_2016[features_in_model[:-1]]\n",
    "#X.sample(10)\n",
    "del features_full_2016\n",
    "\n",
    "# impute using the mean as a simple solution\n",
    "for col in X_full.columns.values:\n",
    "    if sum(X_full[col].isna()) > 0: # only apply this to columns with missing data\n",
    "        imputer.fit(X_full[[col]])\n",
    "        X_full[col] = imputer.transform(X_full[[col]]).ravel()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# now produce predictions for the full dataset using the trained model\n",
    "Y_pred_full = gbt_train.predict(X_full)\n",
    "\n",
    "Y_pred_df = pd.DataFrame({'ParcelId': X_full.index, 'y_pred' : Y_pred_full})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "## write output file for submission to kaggle \n",
    "#\n",
    "#file = open(fname_sub, 'w')\n",
    "#file.write('ParcelId,201610,201611,201612,201710,201711,201712\\n')\n",
    "#for i in range(len(Y_pred_df)):\n",
    "##    print(Y_pred_df.ParcelId[i], Y_pred_df.y_pred.values[i])\n",
    "#    file.write('%u,%4.4f,%4.4f,%4.4f,%4.4f,%4.4f,%4.4f\\n' % (\n",
    "#        Y_pred_df.ParcelId[i], \n",
    "#        Y_pred_df.y_pred.values[i], Y_pred_df.y_pred.values[i], Y_pred_df.y_pred.values[i], \n",
    "#        Y_pred_df.y_pred.values[i], Y_pred_df.y_pred.values[i], Y_pred_df.y_pred.values[i]))\n",
    "#file.close()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:insight]",
   "language": "python",
   "name": "conda-env-insight-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}