04.09-Mean-Median-Imputation-Sklearn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Mean / median imputation with Scikit-learn ==> SimpleImputer\n",
    "\n",
    "Scikit-learn provides a class to make most of the most common data imputation techniques.\n",
    "\n",
    "The **SimpleImputer** class provides basic strategies for imputing missing values, including:\n",
    "\n",
    "- Mean and median imputation for numerical variables\n",
    "- Most frequent category imputation for categorical variables\n",
    "- Arbitrary value imputation for both categorical and numerical variables\n",
    "\n",
    "### Advantages\n",
    "\n",
    "- Simple to use if applied to the entire dataframe\n",
    "- Maintained by the Scikit-learn developers: good quality code\n",
    "- Fast computation (it uses numpy for calculations)\n",
    "- Allows for grid search over the various imputation techniques\n",
    "- Allows for different missing values encodings (you can indicate if the missing values are np.nan, or zeroes, etc)\n",
    "\n",
    "### Limitations\n",
    "\n",
    "- Returns a numpy array instead of a pandas dataframe, inconvenient for data analysis\n",
    "- Needs to use additional classes to select which features to impute  ==>\n",
    "    - requires more lines of code\n",
    "    - additional classes still in beta (may change without warning)\n",
    "    - not so straightforward to use anymore.\n",
    "\n",
    "### More details about the transformers\n",
    "\n",
    "- [SimpleImputer](https://scikit-learn.org/stable/modules/generated/sklearn.impute.SimpleImputer.html#sklearn.impute.SimpleImputer)\n",
    "- [ColumnTransformer](https://scikit-learn.org/stable/modules/generated/sklearn.compose.ColumnTransformer.html)\n",
    "- [Stackoverflow](https://stackoverflow.com/questions/54160370/how-to-use-sklearn-column-transformer)\n",
    "\n",
    "\n",
    "## In this demo:\n",
    "\n",
    "We will do **mean / median imputation with Scikit-learn** using the Ames House Price Dataset.\n",
    "\n",
    "- To download the dataset please refer to the lecture **Datasets** in **Section 1** of this course."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# these are the objects we need to impute missing data\n",
    "# with sklearn\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.pipeline import Pipeline\n",
    "\n",
    "# to split the datasets\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we use only the following variables for the demo:\n",
    "# 3 of which contain NA\n",
    "\n",
    "cols_to_use = [\n",
    "    'OverallQual', 'TotalBsmtSF', '1stFlrSF', 'GrLivArea', 'WoodDeckSF',\n",
    "    'BsmtUnfSF', 'LotFrontage', 'MasVnrArea', 'GarageYrBlt', 'SalePrice'\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1460, 10)\n"
     ]
    },
    {
     "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>LotFrontage</th>\n",
       "      <th>OverallQual</th>\n",
       "      <th>MasVnrArea</th>\n",
       "      <th>BsmtUnfSF</th>\n",
       "      <th>TotalBsmtSF</th>\n",
       "      <th>1stFlrSF</th>\n",
       "      <th>GrLivArea</th>\n",
       "      <th>GarageYrBlt</th>\n",
       "      <th>WoodDeckSF</th>\n",
       "      <th>SalePrice</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>65.0</td>\n",
       "      <td>7</td>\n",
       "      <td>196.0</td>\n",
       "      <td>150</td>\n",
       "      <td>856</td>\n",
       "      <td>856</td>\n",
       "      <td>1710</td>\n",
       "      <td>2003.0</td>\n",
       "      <td>0</td>\n",
       "      <td>208500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>80.0</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>284</td>\n",
       "      <td>1262</td>\n",
       "      <td>1262</td>\n",
       "      <td>1262</td>\n",
       "      <td>1976.0</td>\n",
       "      <td>298</td>\n",
       "      <td>181500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>68.0</td>\n",
       "      <td>7</td>\n",
       "      <td>162.0</td>\n",
       "      <td>434</td>\n",
       "      <td>920</td>\n",
       "      <td>920</td>\n",
       "      <td>1786</td>\n",
       "      <td>2001.0</td>\n",
       "      <td>0</td>\n",
       "      <td>223500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>60.0</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>540</td>\n",
       "      <td>756</td>\n",
       "      <td>961</td>\n",
       "      <td>1717</td>\n",
       "      <td>1998.0</td>\n",
       "      <td>0</td>\n",
       "      <td>140000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>84.0</td>\n",
       "      <td>8</td>\n",
       "      <td>350.0</td>\n",
       "      <td>490</td>\n",
       "      <td>1145</td>\n",
       "      <td>1145</td>\n",
       "      <td>2198</td>\n",
       "      <td>2000.0</td>\n",
       "      <td>192</td>\n",
       "      <td>250000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   LotFrontage  OverallQual  MasVnrArea  BsmtUnfSF  TotalBsmtSF  1stFlrSF  \\\n",
       "0         65.0            7       196.0        150          856       856   \n",
       "1         80.0            6         0.0        284         1262      1262   \n",
       "2         68.0            7       162.0        434          920       920   \n",
       "3         60.0            7         0.0        540          756       961   \n",
       "4         84.0            8       350.0        490         1145      1145   \n",
       "\n",
       "   GrLivArea  GarageYrBlt  WoodDeckSF  SalePrice  \n",
       "0       1710       2003.0           0     208500  \n",
       "1       1262       1976.0         298     181500  \n",
       "2       1786       2001.0           0     223500  \n",
       "3       1717       1998.0           0     140000  \n",
       "4       2198       2000.0         192     250000  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's load the House Prices dataset\n",
    "\n",
    "data = pd.read_csv('../houseprice.csv', usecols=cols_to_use)\n",
    "print(data.shape)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LotFrontage    0.177397\n",
       "OverallQual    0.000000\n",
       "MasVnrArea     0.005479\n",
       "BsmtUnfSF      0.000000\n",
       "TotalBsmtSF    0.000000\n",
       "1stFlrSF       0.000000\n",
       "GrLivArea      0.000000\n",
       "GarageYrBlt    0.055479\n",
       "WoodDeckSF     0.000000\n",
       "SalePrice      0.000000\n",
       "dtype: float64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's check the null values\n",
    "data.isnull().mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The variables LotFrontage, MasVnrArea and GarageYrBlt contain missing data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1022, 9), (438, 9))"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's separate into training and testing set\n",
    "\n",
    "# first let's remove the target from the features\n",
    "cols_to_use.remove('SalePrice')\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(data[cols_to_use], # just the features\n",
    "                                                    data['SalePrice'], # the target\n",
    "                                                    test_size=0.3, # the percentage of obs in the test set\n",
    "                                                    random_state=0) # for reproducibility\n",
    "X_train.shape, X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OverallQual    0.000000\n",
       "TotalBsmtSF    0.000000\n",
       "1stFlrSF       0.000000\n",
       "GrLivArea      0.000000\n",
       "WoodDeckSF     0.000000\n",
       "BsmtUnfSF      0.000000\n",
       "LotFrontage    0.184932\n",
       "MasVnrArea     0.004892\n",
       "GarageYrBlt    0.052838\n",
       "dtype: float64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's check the misssing data again\n",
    "X_train.isnull().mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### SimpleImputer on the entire dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SimpleImputer(copy=True, fill_value=None, missing_values=nan,\n",
       "       strategy='median', verbose=0)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Now we impute the missing values with SimpleImputer\n",
    "\n",
    "# create an instance of the simple imputer\n",
    "# we indicate that we want to impute with the median\n",
    "imputer = SimpleImputer(strategy='median')\n",
    "\n",
    "# we fit the imputer to the train set\n",
    "# the imputer will learn the median of all variables\n",
    "imputer.fit(X_train[cols_to_use])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Note, if you want to do mean imputation instead of median, pass the argument 'mean' to 'strategy'**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   6. ,  992. , 1095. , 1479. ,    0. ,  486.5,   69. ,    0. ,\n",
       "       1979. ])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# we can look at the learnt medians like this:\n",
    "imputer.statistics_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OverallQual       6.0\n",
       "TotalBsmtSF     992.0\n",
       "1stFlrSF       1095.0\n",
       "GrLivArea      1479.0\n",
       "WoodDeckSF        0.0\n",
       "BsmtUnfSF       486.5\n",
       "LotFrontage      69.0\n",
       "MasVnrArea        0.0\n",
       "GarageYrBlt    1979.0\n",
       "dtype: float64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and we can investigate the medians directly to corroborate\n",
    "# the imputer did a good job\n",
    "X_train[cols_to_use].median()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Note** that SimpleImputer learns the medians of ALL the variables in the dataset, those with NA and those without NA."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[   7., 1057., 1057., ...,   69.,  573., 1998.],\n",
       "       [   6., 1291., 1291., ...,   69.,    0., 1996.],\n",
       "       [   5.,  858.,  858., ...,   50.,    0., 1979.],\n",
       "       ...,\n",
       "       [   6.,    0., 1318., ...,   68.,    0., 1978.],\n",
       "       [   7., 1374., 1557., ...,   69.,   18., 2003.],\n",
       "       [   7., 1195., 1195., ...,   58.,   30., 1998.]])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and now we impute the train and test set\n",
    "\n",
    "# NOTE: the data is returned as a numpy array!!!\n",
    "X_train = imputer.transform(X_train)\n",
    "X_test = imputer.transform(X_test)\n",
    "\n",
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "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>OverallQual</th>\n",
       "      <th>TotalBsmtSF</th>\n",
       "      <th>1stFlrSF</th>\n",
       "      <th>GrLivArea</th>\n",
       "      <th>WoodDeckSF</th>\n",
       "      <th>BsmtUnfSF</th>\n",
       "      <th>LotFrontage</th>\n",
       "      <th>MasVnrArea</th>\n",
       "      <th>GarageYrBlt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7.0</td>\n",
       "      <td>1057.0</td>\n",
       "      <td>1057.0</td>\n",
       "      <td>2034.0</td>\n",
       "      <td>576.0</td>\n",
       "      <td>318.0</td>\n",
       "      <td>69.0</td>\n",
       "      <td>573.0</td>\n",
       "      <td>1998.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>307.0</td>\n",
       "      <td>288.0</td>\n",
       "      <td>69.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1996.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>5.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>117.0</td>\n",
       "      <td>162.0</td>\n",
       "      <td>50.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1979.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6.0</td>\n",
       "      <td>560.0</td>\n",
       "      <td>698.0</td>\n",
       "      <td>1258.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>356.0</td>\n",
       "      <td>60.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1939.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2.0</td>\n",
       "      <td>290.0</td>\n",
       "      <td>438.0</td>\n",
       "      <td>438.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>60.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1930.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   OverallQual  TotalBsmtSF  1stFlrSF  GrLivArea  WoodDeckSF  BsmtUnfSF  \\\n",
       "0          7.0       1057.0    1057.0     2034.0       576.0      318.0   \n",
       "1          6.0       1291.0    1291.0     1291.0       307.0      288.0   \n",
       "2          5.0        858.0     858.0      858.0       117.0      162.0   \n",
       "3          6.0        560.0     698.0     1258.0         0.0      356.0   \n",
       "4          2.0        290.0     438.0      438.0         0.0        0.0   \n",
       "\n",
       "   LotFrontage  MasVnrArea  GarageYrBlt  \n",
       "0         69.0       573.0       1998.0  \n",
       "1         69.0         0.0       1996.0  \n",
       "2         50.0         0.0       1979.0  \n",
       "3         60.0         0.0       1939.0  \n",
       "4         60.0         0.0       1930.0  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# if we wanted to continue our data analysis, we would have to\n",
    "# encode the train set back to a dataframe:\n",
    "\n",
    "pd.DataFrame(X_train, columns=cols_to_use).head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Let's explore the change in the distributions after the median imputation\n",
    "\n",
    "X_train = pd.DataFrame(X_train, columns=cols_to_use)\n",
    "X_train.hist(bins=50, figsize=(10,10))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note in the variable LotFrontAge the tall bar at the value of the median in the histogram, which appears after the data imputation.\n",
    "\n",
    "**A MASSIVE NOTE OF CAUTION**: when encoding the returned numpy arrays from Scikit-learn back to a dataframe, there is the potential risk of messing the order of the features and therefore MISS-REPRESENTING the data!!\n",
    "\n",
    "You need to be extremely careful to pass the columns to the pd.DataFrame command in the same order in which they were in the X_train BEFORE the transformation!.\n",
    "\n",
    "\n",
    "### SimpleImputer: different procedures on different features\n",
    "\n",
    "On occasions, we want to implement different imputation techniques to different features, how can we do this using the SimpleImputer from Scikit-learn?\n",
    "\n",
    "We need to use another class called ColumnTransformer. \n",
    "\n",
    "**Note** ColumnTransformer has been recently released by Scikit-learn and it is therefore in beta version. This means that the developers can change the way it is used, or remove it altogether any time. The implication is that the code you write today, may not be useful after new releases from Scikit-learn. That is usually what happens with any open-source package when the developers are trying new functionality.\n",
    "\n",
    "More details about the ColumnTransformer in the [Sklearn website](https://scikit-learn.org/stable/modules/generated/sklearn.compose.ColumnTransformer.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1022, 9), (438, 9))"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's separate into training and testing set\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(data[cols_to_use],# predictors only\n",
    "                                                    data['SalePrice'],\n",
    "                                                    test_size=0.3,\n",
    "                                                    random_state=0)\n",
    "X_train.shape, X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OverallQual    0.000000\n",
       "TotalBsmtSF    0.000000\n",
       "1stFlrSF       0.000000\n",
       "GrLivArea      0.000000\n",
       "WoodDeckSF     0.000000\n",
       "BsmtUnfSF      0.000000\n",
       "LotFrontage    0.184932\n",
       "MasVnrArea     0.004892\n",
       "GarageYrBlt    0.052838\n",
       "dtype: float64"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's look at the missing values\n",
    "X_train.isnull().mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# let's look at the distributions of the variables to decide\n",
    "# whether we use the mean or the median\n",
    "\n",
    "X_train.hist(bins=50, figsize=(10,10))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this demo, I will impute LotFrontage with the mean, and MasVnrArea and GarageYrBlt with the median, as those variables are quite skewed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# first we need to make lists, indicating which features\n",
    "# will be imputed with each method\n",
    "\n",
    "numeric_features_mean = ['LotFrontage']\n",
    "numeric_features_median = ['MasVnrArea', 'GarageYrBlt']\n",
    "\n",
    "# then we instantiate the imputers, within a pipeline\n",
    "# we create one mean imputer and one median imputer\n",
    "# by changing the parameter in the strategy\n",
    "\n",
    "numeric_mean_imputer = Pipeline(steps=[\n",
    "    ('imputer', SimpleImputer(strategy='mean')),\n",
    "])\n",
    "\n",
    "numeric_median_imputer = Pipeline(steps=[\n",
    "    ('imputer', SimpleImputer(strategy='median')),\n",
    "])\n",
    "\n",
    "# then we put the features list and the transformers together\n",
    "# using the column transformer\n",
    "\n",
    "# we need to add remainder = True to indicate what we want\n",
    "# ALL the columns returned at the end of the transformation\n",
    "# and not just the engineered ones, which is the default\n",
    "# behaviour of ColumnTransformer. \n",
    "\n",
    "preprocessor = ColumnTransformer(transformers=[\n",
    "    ('mean_imputer', numeric_mean_imputer, numeric_features_mean),\n",
    "    ('median_imputer', numeric_median_imputer, numeric_features_median)\n",
    "], remainder='passthrough')\n",
    "\n",
    "\n",
    "# parameters of the ColumnTransformer\n",
    "# remainder = 'passthrough' indicates that we want to retain ALL the columns in the dataset\n",
    "            # otherwise only those specified in the imputing steps will be kept\n",
    "    \n",
    "# for more details follow the sklearn page:\n",
    "# https://scikit-learn.org/stable/modules/generated/sklearn.compose.ColumnTransformer.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ColumnTransformer(n_jobs=None, remainder='passthrough', sparse_threshold=0.3,\n",
       "         transformer_weights=None,\n",
       "         transformers=[('mean_imputer', Pipeline(memory=None,\n",
       "     steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan, strategy='mean',\n",
       "       verbose=0))]), ['LotFrontage']), ('median_imputer', Pipeline(memory=None,\n",
       "     steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan,\n",
       "       strategy='median', verbose=0))]), ['MasVnrArea', 'GarageYrBlt'])])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# now we fit the preprocessor\n",
    "preprocessor.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('mean_imputer', Pipeline(memory=None,\n",
       "       steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan, strategy='mean',\n",
       "         verbose=0))]), ['LotFrontage']),\n",
       " ('median_imputer', Pipeline(memory=None,\n",
       "       steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan,\n",
       "         strategy='median', verbose=0))]), ['MasVnrArea', 'GarageYrBlt'])]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# we can explore the transformers like this:\n",
    "\n",
    "preprocessor.transformers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([69.66866747])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and we can look at the parameters learnt like this:\n",
    "\n",
    "# for the mean imputer\n",
    "preprocessor.named_transformers_['mean_imputer'].named_steps['imputer'].statistics_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LotFrontage    69.668667\n",
       "dtype: float64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and we can corroborate the value with that one in\n",
    "# the train set\n",
    "X_train[numeric_features_mean].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   0., 1979.])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# for the median imputerthere are 2 values learnt\n",
    "# because we want to impute 2 different variables\n",
    "\n",
    "preprocessor.named_transformers_['median_imputer'].named_steps['imputer'].statistics_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MasVnrArea        0.0\n",
       "GarageYrBlt    1979.0\n",
       "dtype: float64"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and we corroborate those values in the train set\n",
    "X_train[numeric_features_median].median()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and now we can impute the data\n",
    "X_train = preprocessor.transform(X_train)\n",
    "\n",
    "# and check it worked\n",
    "np.mean(np.isnan(X_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# and now we can impute the test data\n",
    "X_test = preprocessor.transform(X_test)\n",
    "\n",
    "# and check it worked\n",
    "np.mean(np.isnan(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  69.66866747,  573.        , 1998.        , ..., 2034.        ,\n",
       "         576.        ,  318.        ],\n",
       "       [  69.66866747,    0.        , 1996.        , ..., 1291.        ,\n",
       "         307.        ,  288.        ],\n",
       "       [  50.        ,    0.        , 1979.        , ...,  858.        ,\n",
       "         117.        ,  162.        ],\n",
       "       ...,\n",
       "       [  68.        ,    0.        , 1978.        , ..., 1902.        ,\n",
       "           0.        ,    0.        ],\n",
       "       [  69.66866747,   18.        , 2003.        , ..., 1557.        ,\n",
       "         143.        , 1374.        ],\n",
       "       [  58.        ,   30.        , 1998.        , ..., 1839.        ,\n",
       "           0.        , 1195.        ]])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# remember that the returned object  is a NumPy array\n",
    "X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('mean_imputer', Pipeline(memory=None,\n",
       "       steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan, strategy='mean',\n",
       "         verbose=0))]), ['LotFrontage']),\n",
       " ('median_imputer', Pipeline(memory=None,\n",
       "       steps=[('imputer', SimpleImputer(copy=True, fill_value=None, missing_values=nan,\n",
       "         strategy='median', verbose=0))]), ['MasVnrArea', 'GarageYrBlt']),\n",
       " ('remainder', 'passthrough', [0, 1, 2, 3, 4, 5])]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# if we want to put the array back into a dataframe, we need\n",
    "# to be careful with the column order\n",
    "\n",
    "# like this, we can explore the features used in each transformer\n",
    "# including those in the remainder (the ones that were not transformed)\n",
    "preprocessor.transformers_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "'remainder', 'passthrough', [0, 1, 2, 3, 4, 5]\n",
    "\n",
    "the code above tells us that the remainder columns are those in position 0, 1, 2, 3, 4, 5 of the X_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['OverallQual',\n",
       " 'TotalBsmtSF',\n",
       " '1stFlrSF',\n",
       " 'GrLivArea',\n",
       " 'WoodDeckSF',\n",
       " 'BsmtUnfSF']"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's capture the columns in a list\n",
    "\n",
    "remainder_cols = [cols_to_use[c] for c in [0, 1, 2, 3, 4, 5]]\n",
    "remainder_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "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>LotFrontage</th>\n",
       "      <th>MasVnrArea</th>\n",
       "      <th>GarageYrBlt</th>\n",
       "      <th>OverallQual</th>\n",
       "      <th>TotalBsmtSF</th>\n",
       "      <th>1stFlrSF</th>\n",
       "      <th>GrLivArea</th>\n",
       "      <th>WoodDeckSF</th>\n",
       "      <th>BsmtUnfSF</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>69.668667</td>\n",
       "      <td>573.0</td>\n",
       "      <td>1998.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>1057.0</td>\n",
       "      <td>1057.0</td>\n",
       "      <td>2034.0</td>\n",
       "      <td>576.0</td>\n",
       "      <td>318.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>69.668667</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1996.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>1291.0</td>\n",
       "      <td>307.0</td>\n",
       "      <td>288.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>50.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1979.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>858.0</td>\n",
       "      <td>117.0</td>\n",
       "      <td>162.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>60.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1939.0</td>\n",
       "      <td>6.0</td>\n",
       "      <td>560.0</td>\n",
       "      <td>698.0</td>\n",
       "      <td>1258.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>356.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>60.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1930.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>290.0</td>\n",
       "      <td>438.0</td>\n",
       "      <td>438.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   LotFrontage  MasVnrArea  GarageYrBlt  OverallQual  TotalBsmtSF  1stFlrSF  \\\n",
       "0    69.668667       573.0       1998.0          7.0       1057.0    1057.0   \n",
       "1    69.668667         0.0       1996.0          6.0       1291.0    1291.0   \n",
       "2    50.000000         0.0       1979.0          5.0        858.0     858.0   \n",
       "3    60.000000         0.0       1939.0          6.0        560.0     698.0   \n",
       "4    60.000000         0.0       1930.0          2.0        290.0     438.0   \n",
       "\n",
       "   GrLivArea  WoodDeckSF  BsmtUnfSF  \n",
       "0     2034.0       576.0      318.0  \n",
       "1     1291.0       307.0      288.0  \n",
       "2      858.0       117.0      162.0  \n",
       "3     1258.0         0.0      356.0  \n",
       "4      438.0         0.0        0.0  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# capture the data back in a dataframe\n",
    "pd.DataFrame(X_train,\n",
    "             columns = numeric_features_mean+numeric_features_median+remainder_cols).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**ANOTHER NOTE OF CAUTION**\n",
    "\n",
    "ColumnTransformer is in essence just slicing the dataframe in the required features. Then the SimpleImputer operates on the sliced dataframe. If different features are required by the different imputation methods, it will slice the dataframe on the required features for each imputation, and then concatenate them to produce the final output.\n",
    "\n",
    "For our specific example, what happens is the following:\n",
    "\n",
    "- ColumnTransformer first creates a single column dataframe with the first feature, LotFrontage\n",
    "- ColumnTransformer creates a 2 column dataframe with the features MasVnrArea and GarageBlt\n",
    "- ColumnTransformer creates an additional dataframe with the remaining features in the train set\n",
    "- The SimpleImputers learn the parameters in their respective sliced dataframes\n",
    "- ColunmnTransformer contatenates back all the dataframes into the final result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "feml",
   "language": "python",
   "name": "feml"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}