Kaggle_Titanic.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# **Kaggle / Titanic, Machine Learning from Disaster**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## **Tutorial competition description**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From [Kaggle](https://www.kaggle.com/c/titanic) :\n",
    "<br> <br>\n",
    "*The sinking of the RMS Titanic is one of the most infamous shipwrecks in history.  On April 15, 1912, during her maiden voyage, the Titanic sank after colliding with an iceberg, killing 1502 out of 2224 passengers and crew. This sensational tragedy shocked the international community and led to better safety regulations for ships.*\n",
    "\n",
    "*One of the reasons that the shipwreck led to such loss of life was that there were not enough lifeboats for the passengers and crew. Although there was some element of luck involved in surviving the sinking, some groups of people were more likely to survive than others, such as women, children, and the upper-class.*\n",
    "\n",
    "*In this challenge, we ask you to complete the analysis of what sorts of people were likely to survive. In particular, we ask you to apply the tools of machine learning to predict which passengers survived the tragedy.*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *Key skills* : Exploratory data analysis, Data cleaning, Binary Classification, Ensemble methods"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Set up**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Importing librairies**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys, os\n",
    "import re\n",
    "from six.moves import urllib\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from pandas.core.common import SettingWithCopyWarning\n",
    "import re\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from IPython.display import display #to have nice dataframe in Jupyter notebook\n",
    "\n",
    "import seaborn as sns\n",
    "import itertools\n",
    "import sklearn\n",
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import StandardScaler, OneHotEncoder\n",
    "from sklearn.impute import SimpleImputer\n",
    "from sklearn.compose import ColumnTransformer\n",
    "from sklearn.svm import SVC, LinearSVC\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.ensemble import RandomForestClassifier, ExtraTreesClassifier, VotingClassifier, AdaBoostClassifier\n",
    "from sklearn.model_selection import cross_val_score, GridSearchCV, RandomizedSearchCV, cross_val_predict\n",
    "from scipy.stats import uniform, randint, expon, reciprocal\n",
    "from sklearn.externals import joblib\n",
    "from statsmodels.distributions.empirical_distribution import ECDF\n",
    "import xgboost as xgb\n",
    "\n",
    "import warnings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Python interpreter :  3.7.3 (default, Apr 24 2019, 15:29:51) [MSC v.1915 64 bit (AMD64)]\n",
      "Numpy version :  1.16.3\n",
      "Pandas version :  0.24.2\n",
      "Matplotlib version :  3.0.3\n",
      "Seaborn version :  0.9.0\n",
      "Scikit learn version :  0.20.3\n",
      "XGboost version :  0.82\n"
     ]
    }
   ],
   "source": [
    "print(\"Python interpreter : \", sys.version)\n",
    "print(\"Numpy version : \", np.__version__)\n",
    "print(\"Pandas version : \", pd.__version__)\n",
    "print(\"Matplotlib version : \", matplotlib.__version__)\n",
    "print(\"Seaborn version : \", sns.__version__)\n",
    "print(\"Scikit learn version : \", sklearn.__version__)\n",
    "print(\"XGboost version : \", xgb.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "warnings.simplefilter(action='ignore', category=(FutureWarning, SettingWithCopyWarning))\n",
    "plt.style.use('ggplot')\n",
    "pd.set_option('display.max_columns', 100)\n",
    "URL_ROOT_PATH = 'https://raw.githubusercontent.com/GuillaumeHarel/Hello_world_ML/master/Raw_data/Kaggle_titanic/'\n",
    "FILE_NAMES = ['train.csv', 'test.csv']\n",
    "MODEL_SAVING_PATH = \"./Models/\"\n",
    "DATA_SAVING_PATH = \"./Data/\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Downloading the data**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def fetch_datasets(dl_root = URL_ROOT_PATH, file_path=DATA_SAVING_PATH):\n",
    "    datasets = {}\n",
    "    if os.path.exists(file_path):\n",
    "        print(\"Files were previously downloaded, loading file from: \", file_path)\n",
    "        for root, dirs, files in os.walk(file_path):\n",
    "            for file in files:\n",
    "                datasets[file] = pd.read_csv(os.path.join(root, file), index_col=0)\n",
    "        return datasets\n",
    "    os.makedirs(file_path)\n",
    "    for file_name in FILE_NAMES:\n",
    "        dl_url = URL_ROOT_PATH + file_name\n",
    "        file_path = DATA_SAVING_PATH + file_name\n",
    "        urllib.request.urlretrieve(dl_url, file_path)\n",
    "        print(\"{} has been successfully donwloaded and saved in {}\".format(file_name, file_path))\n",
    "        print('~-'*30)\n",
    "        datasets[file_name] = pd.read_csv(file_path, index_col=0)\n",
    "    return datasets\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files were previously downloaded, loading file from:  ./Data/\n"
     ]
    }
   ],
   "source": [
    "datasets = fetch_datasets()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['test.csv', 'train.csv'])\n",
      "(891, 11)\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>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>1</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>7.2500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>71.2833</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>7.9250</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>1</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>7.2500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>71.2833</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>7.9250</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(418, 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>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>892</th>\n",
       "      <td>3</td>\n",
       "      <td>Kelly, Mr. James</td>\n",
       "      <td>male</td>\n",
       "      <td>34.5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>330911</td>\n",
       "      <td>7.8292</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>893</th>\n",
       "      <td>3</td>\n",
       "      <td>Wilkes, Mrs. James (Ellen Needs)</td>\n",
       "      <td>female</td>\n",
       "      <td>47.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>363272</td>\n",
       "      <td>7.0000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>894</th>\n",
       "      <td>2</td>\n",
       "      <td>Myles, Mr. Thomas Francis</td>\n",
       "      <td>male</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>240276</td>\n",
       "      <td>9.6875</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>892</th>\n",
       "      <td>3</td>\n",
       "      <td>Kelly, Mr. James</td>\n",
       "      <td>male</td>\n",
       "      <td>34.5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>330911</td>\n",
       "      <td>7.8292</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>893</th>\n",
       "      <td>3</td>\n",
       "      <td>Wilkes, Mrs. James (Ellen Needs)</td>\n",
       "      <td>female</td>\n",
       "      <td>47.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>363272</td>\n",
       "      <td>7.0000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>894</th>\n",
       "      <td>2</td>\n",
       "      <td>Myles, Mr. Thomas Francis</td>\n",
       "      <td>male</td>\n",
       "      <td>62.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>240276</td>\n",
       "      <td>9.6875</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(datasets.keys())\n",
    "for file in FILE_NAMES:\n",
    "    print(datasets[file].shape)\n",
    "    display(datasets[file].head(3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Exploratory Data Analysis**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 891 entries, 1 to 891\n",
      "Data columns (total 11 columns):\n",
      "Survived    891 non-null int64\n",
      "Pclass      891 non-null int64\n",
      "Name        891 non-null object\n",
      "Sex         891 non-null object\n",
      "Age         714 non-null float64\n",
      "SibSp       891 non-null int64\n",
      "Parch       891 non-null int64\n",
      "Ticket      891 non-null object\n",
      "Fare        891 non-null float64\n",
      "Cabin       204 non-null object\n",
      "Embarked    889 non-null object\n",
      "dtypes: float64(2), int64(4), object(5)\n",
      "memory usage: 83.5+ KB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "df_train = datasets['train.csv']\n",
    "df_test = datasets['test.csv']\n",
    "print(df_train.info())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Object dtype corresponds to strings while int64 et float64 object respectively stands for numeric categorical features and numeric continuous features. Survived viariable is the target variable we want to predict for the test dataset. \n",
    "Meaning of each feature is quickly indicated below :\n",
    "\n",
    "\n",
    "    Survived: that's the target, 0 means the passenger did not survive, while 1 means he/she survived.\n",
    "    Pclass: passenger class.\n",
    "    Name, Sex, Age: self-explanatory\n",
    "    SibSp: how many siblings & spouses of the passenger aboard the Titanic.\n",
    "    Parch: how many children & parents of the passenger aboard the Titanic.\n",
    "    Ticket: ticket id\n",
    "    Fare: price paid (in pounds)\n",
    "    Cabin: passenger's cabin number\n",
    "    Embarked: where the passenger embarked the Titanic\n",
    "\n",
    "\n",
    "\n",
    "From the .info method, some missing values (NaN values) can be observed for the Age (about roughly 20%) and for the Cabin number (> 75%). Two values are also missing for the Embarked feature. For these 3 features, missing values will have to be inmputed in order to be included in the model. Let's have a closer look at the train df head."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>1</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>7.2500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>71.2833</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>7.9250</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Futrelle, Mrs. Jacques Heath (Lily May Peel)</td>\n",
       "      <td>female</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>113803</td>\n",
       "      <td>53.1000</td>\n",
       "      <td>C123</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Allen, Mr. William Henry</td>\n",
       "      <td>male</td>\n",
       "      <td>35.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>373450</td>\n",
       "      <td>8.0500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Moran, Mr. James</td>\n",
       "      <td>male</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>330877</td>\n",
       "      <td>8.4583</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>McCarthy, Mr. Timothy J</td>\n",
       "      <td>male</td>\n",
       "      <td>54.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>17463</td>\n",
       "      <td>51.8625</td>\n",
       "      <td>E46</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Palsson, Master. Gosta Leonard</td>\n",
       "      <td>male</td>\n",
       "      <td>2.0</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>349909</td>\n",
       "      <td>21.0750</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Johnson, Mrs. Oscar W (Elisabeth Vilhelmina Berg)</td>\n",
       "      <td>female</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>347742</td>\n",
       "      <td>11.1333</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>Nasser, Mrs. Nicholas (Adele Achem)</td>\n",
       "      <td>female</td>\n",
       "      <td>14.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>237736</td>\n",
       "      <td>30.0708</td>\n",
       "      <td>NaN</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PassengerId</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>1</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>7.2500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>71.2833</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>7.9250</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Futrelle, Mrs. Jacques Heath (Lily May Peel)</td>\n",
       "      <td>female</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>113803</td>\n",
       "      <td>53.1000</td>\n",
       "      <td>C123</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Allen, Mr. William Henry</td>\n",
       "      <td>male</td>\n",
       "      <td>35.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>373450</td>\n",
       "      <td>8.0500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Moran, Mr. James</td>\n",
       "      <td>male</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>330877</td>\n",
       "      <td>8.4583</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>McCarthy, Mr. Timothy J</td>\n",
       "      <td>male</td>\n",
       "      <td>54.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>17463</td>\n",
       "      <td>51.8625</td>\n",
       "      <td>E46</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Palsson, Master. Gosta Leonard</td>\n",
       "      <td>male</td>\n",
       "      <td>2.0</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>349909</td>\n",
       "      <td>21.0750</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Johnson, Mrs. Oscar W (Elisabeth Vilhelmina Berg)</td>\n",
       "      <td>female</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>347742</td>\n",
       "      <td>11.1333</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>Nasser, Mrs. Nicholas (Adele Achem)</td>\n",
       "      <td>female</td>\n",
       "      <td>14.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>237736</td>\n",
       "      <td>30.0708</td>\n",
       "      <td>NaN</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "display(df_train.head(10))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Cabin number feature might be a relevant feature since it might provide useful information on cabin location (e.g. top deck) but the low proportion of available data is trully an issue since missing values can hardly be imputed from other features. I will be likely to drop this feature.\n",
    "<br> <br>\n",
    "About the string features :\n",
    "<br>\n",
    "Ticket feature does not seem very useful at first sight. Although it might contain information (e.g. \"PC\" must stands for Premiere Class), most of them appear redundant with other features and the information format seems to vary a lot. The only useful information I could think off would be to link passenger between themselves (provided group of people can travel with a same ticket ?).\n",
    "<br>\n",
    "Name feature present a few information that might be useful like the name (to link family member/spouse) or a specific \"title\" (eg. \"Miss\") bringing extra information on the social status or age.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's quickly go through continuous and categorical feature before digging into EDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total nb of passengers in the dataset: 1309 VS 2224 passengers + crew reported in records\n",
      "Mean survival rate - train set: 38.4% VS real records : 32.5%\n"
     ]
    }
   ],
   "source": [
    "th_total_pass_nb = 2224\n",
    "th_total_death = 1502\n",
    "th_SR = (th_total_pass_nb - th_total_death)/th_total_pass_nb\n",
    "data_total_pass_nb = len(df_train) + len(df_test)\n",
    "mean_train_SR = df_train['Survived'].mean()\n",
    "print(\"Total nb of passengers in the dataset: {} VS {} passengers + crew reported in records\".format(data_total_pass_nb, th_total_pass_nb))\n",
    "print(\"Mean survival rate - train set: {:.1f}% VS real records : {:.1f}%\".format(mean_train_SR*100, th_SR*100))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here I guess that many of \"missing\" passengers in the dataset correspond to crew members. Some passengers must probably still have been discarded from the dataset because of poor/unavailable data. At some point I think I read somewhere that the list of passengers and corresponding information were issued from an hand-written embarkation registry partially found after the shipwreck."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>Age</th>\n",
       "      <th>Fare</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>714.000000</td>\n",
       "      <td>891.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>29.699118</td>\n",
       "      <td>32.204208</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>14.526497</td>\n",
       "      <td>49.693429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.420000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>20.125000</td>\n",
       "      <td>7.910400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>28.000000</td>\n",
       "      <td>14.454200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>38.000000</td>\n",
       "      <td>31.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>80.000000</td>\n",
       "      <td>512.329200</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>Age</th>\n",
       "      <th>Fare</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>714.000000</td>\n",
       "      <td>891.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>29.699118</td>\n",
       "      <td>32.204208</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>14.526497</td>\n",
       "      <td>49.693429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.420000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>20.125000</td>\n",
       "      <td>7.910400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>28.000000</td>\n",
       "      <td>14.454200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>38.000000</td>\n",
       "      <td>31.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>80.000000</td>\n",
       "      <td>512.329200</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "display(df_train[['Age', 'Fare']].describe())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    549\n",
       "1    342\n",
       "Name: Survived, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "3    491\n",
       "1    216\n",
       "2    184\n",
       "Name: Pclass, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "male      577\n",
       "female    314\n",
       "Name: Sex, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "0    608\n",
       "1    209\n",
       "2     28\n",
       "4     18\n",
       "3     16\n",
       "8      7\n",
       "5      5\n",
       "Name: SibSp, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "0    678\n",
       "1    118\n",
       "2     80\n",
       "5      5\n",
       "3      5\n",
       "4      4\n",
       "6      1\n",
       "Name: Parch, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "S    644\n",
       "C    168\n",
       "Q     77\n",
       "Name: Embarked, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cat_features = ['Survived', 'Pclass', 'Sex', 'SibSp', 'Parch', 'Embarked']\n",
    "for feat in cat_features:\n",
    "    display(df_train[feat].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "347082          7\n",
       "CA. 2343        7\n",
       "1601            7\n",
       "CA 2144         6\n",
       "347088          6\n",
       "3101295         6\n",
       "382652          5\n",
       "S.O.C. 14879    5\n",
       "4133            4\n",
       "PC 17757        4\n",
       "Name: Ticket, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "display(df_train['Ticket'].value_counts()[:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Thus a single ticket can be used for a group of passenger traveling together."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Studying Survival rate through categorical features**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Survival Rate VS Sex, Embarked and Pclass*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x360 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axarr = plt.subplots(1,3, sharey=True, figsize=(13, 5))\n",
    "sns.pointplot(x='Sex', y='Survived', data=df_train, ax=axarr[0])\n",
    "axarr[0].set(ylabel='Survival rate (mean +/- 95% CI)', xlabel='')\n",
    "sns.catplot(x='Sex', y='Survived', hue='Pclass', data=df_train, ax=axarr[1], kind='bar')\n",
    "axarr[1].tick_params(axis='y', which='both', left=False, right=False, labelleft=False)\n",
    "axarr[1].set(ylabel='', xlabel='')\n",
    "sns.catplot(x='Sex', y='Survived', hue='Embarked', data=df_train, ax=axarr[2], kind='bar')\n",
    "axarr[2].tick_params(axis='y', which='both', left=False, right=True, labelleft=False, labelright=True)\n",
    "axarr[2].set(ylabel='', xlabel='')\n",
    "plt.close(2)\n",
    "plt.close(3)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There is a huge statistically significant difference for the mean survival rate between men and women (roughly 20% VS 80 %!). As a result, the graphical EDA will be conducted through this dichotomy. <br>\n",
    "The PClass seems to also play an important role for the mean survival rate for both men and women. <br>\n",
    "The right graph would suggest that people who embarked at Cherbourg in France were more likely to survive than their american counterpart. I don't see any particular logical reason to that apart from the hypothesis of a specific or 'advantageous' allocation of the cabins for people embarking at Cherbourg. Let's double check with the distribution of embarking dock according to PClass. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1, 1, figsize=(13,5))\n",
    "sns.countplot(x='Embarked', hue='Pclass', data=df_train)\n",
    "plt.close(2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Distribution for embarkation deck along PClass is quite heterogenous... About half of people who embarked at Cherbourg were in Pclass 1, hence the highest survival rate looking back at the previous figure. The Embarked feature might not be very relevant for our model, let's decide that later."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Feature engineering** "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Survival Rate VS Family*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SibSp and Parch are not very convenient features since they kind of mix \"opposite\" attributes into a single (or at least that might have been more useful if separated). Let's combine the whole thing into a single feature called \"Family\" representing the number of people from its family one is traveling with !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x432 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_train['Family'] = df_train[\"SibSp\"] + df_train['Parch']\n",
    "fig, axarr = plt.subplots(2, 3, figsize=(13, 6))\n",
    "for ind, c in enumerate(['SibSp', 'Parch', 'Family']):\n",
    "    ax = axarr[0, ind]\n",
    "    sns.catplot(x=c, y='Survived', data=df_train, ax=ax, kind='bar', hue='Sex')\n",
    "    ax.set(xlabel='', ylabel='Survival rate (mean +/- 95% CI)')\n",
    "    ax.legend(loc='upper right')\n",
    "    ax = axarr[1, ind]\n",
    "    sns.countplot(x=c, data=df_train, hue='Survived', ax=ax)\n",
    "    ax.legend(loc='upper center')\n",
    "plt.close(2)\n",
    "plt.close(3)\n",
    "plt.close(4)\n",
    "plt.tight_layout()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It is quite difficult to appreciate difference in survival rate for SibSp and Parch. Looking at the family we might see that single men seem to present lower survival rates in contrast with men with one or two family member. Above 3 family members it's almost impossible to pronounce oneself because of the low data number and very high variability. Later it would surely be useful to rework the categorical values (e.g. 'single', '1 to 3' and 'above 3'). "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Survival Rate VS Social title*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_train['Title'] = df_train['Name'].map(lambda name: re.split(',.|\\.', name)[1]).astype('category')\n",
    "title_order = df_train.groupby('Title')['Survived'].mean().sort_values(ascending=False).index\n",
    "fig, axarr = plt.subplots(2, 1, figsize=(15, 6))\n",
    "sns.catplot(x='Title', y='Survived', data=df_train, ax=axarr[0], order=title_order, kind='bar')\n",
    "axarr[0].set(xlabel='', ylabel='Survival rate (mean +/- 95% CI)')\n",
    "sns.countplot(x='Title', data=df_train, ax=axarr[1], order=title_order)\n",
    "axarr[1].set_ylim({0,50})\n",
    "axarr[1].set(xlabel='', ylabel='count(zoom 0-50)')\n",
    "plt.close(2)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Mr              517\n",
       "Miss            182\n",
       "Mrs             125\n",
       "Master           40\n",
       "Dr                7\n",
       "Rev               6\n",
       "Col               2\n",
       "Mlle              2\n",
       "Major             2\n",
       "Jonkheer          1\n",
       "Don               1\n",
       "the Countess      1\n",
       "Lady              1\n",
       "Sir               1\n",
       "Mme               1\n",
       "Ms                1\n",
       "Capt              1\n",
       "Name: Title, dtype: int64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "display(df_train[\"Title\"].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Title feature is a bit tricky here because of the very low number of instances for each specific title. Only Miss and Master are well enough represented among the passengers and seem to present an interest in comparison with respectively Mrs and Mr common titles. Later we might want to recode the categories of this feature."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Continuous numeric features**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Survival Rate VS Age and Fare*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axarr = plt.subplots(2, 2, figsize=(12, 6))\n",
    "ecdf_d = {}\n",
    "for idx, c in enumerate(['Age', 'Fare']):\n",
    "    for idy, s in enumerate(df_train.Sex.unique()):\n",
    "        ax = axarr[idx, idy]\n",
    "        for k in df_train.Survived.unique():\n",
    "            ecdf_d['ECDF_{Sex}_{Survived}'.format(Sex=s, Survived=k)] = ECDF(\n",
    "                df_train.loc[(df_train['Sex'] == s) & (df_train['Survived'] == k), c].dropna().values)\n",
    "            ax.plot(ecdf_d['ECDF_{Sex}_{Survived}'.format(Sex=s, Survived=k)].x,\n",
    "                    ecdf_d['ECDF_{Sex}_{Survived}'.format(Sex=s, Survived=k)].y,\n",
    "                    '.', linestyle='none', label='Survived : {} '.format(k), alpha=0.5,\n",
    "                    )\n",
    "            if idx == 0:\n",
    "                ax.set(title='Sex : {} '.format(s), xlabel=c, ylabel='ECDF')\n",
    "            else:\n",
    "                ax.set(xlabel=c, ylabel='ECDF')\n",
    "            ax.legend(loc='best')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ECDFs for Age enlighten different statements according to Sex :\n",
    "\n",
    "- Men : proportion of young boy (below 15) is much more important among survivors. After 25, age distributions appear to be roughly similar between survivors and deceased <br>\n",
    "- Woman : An opposite trend could be observed within women (red curve above blue curve). Women who survived the shipwreck tend to be slighlty older. This is consistent with the small decrease of mean survival rate observed for the \"Miss\" title in comparison with \"Mrs\" title.\n",
    "\n",
    "\n",
    "\n",
    "ECDFs observed for the Fare are consistent with previous statements for Pclass and indicated that people with high fare ticket were more prone to survive. This is particularly true if you consider fare above 50 £.\n",
    "\n",
    "For age, we might later want to convert the continuous numeric feature into categoricical feature. Relying on top ECDFs, bucket 0-15 seems interesting. We could keep the same bucket size for the whole distribution. Let's have a closer look at the mean survival rates along age categories.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Sex     Age_bucket\n",
       "female  0.0           0.615385\n",
       "        15.0          0.750000\n",
       "        30.0          0.807692\n",
       "        45.0          0.781250\n",
       "        60.0          1.000000\n",
       "male    0.0           0.538462\n",
       "        15.0          0.151515\n",
       "        30.0          0.204380\n",
       "        45.0          0.192982\n",
       "        60.0          0.136364\n",
       "Name: Survived, dtype: float64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train['Age_bucket'] = np.clip(df_train['Age']//15 * 15, a_max=60, a_min=None)\n",
    "display(df_train.groupby(by=['Sex', 'Age_bucket'])['Survived'].mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x_tick_lab = ['0-14', '15-29', '30-44', '45-60', '>60']\n",
    "fig, axarr = plt.subplots(1, 2, figsize=(12, 6))\n",
    "sns.catplot(x='Age_bucket', y='Survived', data=df_train.loc[df_train['Sex'] == 'male', ], kind='bar', ax=axarr[0])\n",
    "axarr[0].set(title='Sex : male', xticklabels=x_tick_lab)\n",
    "sns.catplot(x='Age_bucket', y='Survived', data=df_train.loc[df_train['Sex'] == 'female', ], kind='bar', ax=axarr[1])\n",
    "axarr[1].set(title='Sex : female', xticklabels=x_tick_lab)\n",
    "plt.close(2)\n",
    "plt.close(3)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Preprocessing the data**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will need to build a few transformers so we could handle data preprocessing in pipeline ! This concerns feature creation and missing value imputing."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Feature Adder : Family, Title and Age bucket**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "class FeatureAdder(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, add_family=True, relevant_nb=3):\n",
    "        self.add_family = add_family\n",
    "        self.relevant_nb = relevant_nb\n",
    "    def fit(self, df, y=None):\n",
    "        return (self)\n",
    "    def transform(self, df, y=None):\n",
    "        df['Title'] = df['Name'].map(lambda name: re.split(',.|\\.', name)[1])\n",
    "        df.loc[(df['Title'] == 'Ms') | (df['Title'] == 'Mlle'), 'Title'] = 'Miss'\n",
    "        if self.add_family:\n",
    "            df['Family'] = df['SibSp'] + df['Parch']\n",
    "            fam_d = {n: \"small_fam\" if n < self.relevant_nb else \"large_fam\" for n in df['Family'].unique()}\n",
    "            fam_d[0] = 'single'  \n",
    "            df['Family'] = df['Family'].map(fam_d)\n",
    "        return(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "class AgeBucketAdder(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, bucket_size=15, bucket_cap=60):\n",
    "        self.bucket_size = bucket_size\n",
    "        self.bucket_cap = bucket_cap\n",
    "    def fit(self, df, y=None):\n",
    "        return (self)\n",
    "    def transform(self, df, y=None):\n",
    "        df['Age_bucket'] = np.clip(df['Age']//self.bucket_size * self.bucket_size, a_max=self.bucket_cap, a_min=None)\n",
    "        return(df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I need to seperate Age_bucket from other features to add because I will build an imputer for Age missing values based on Title..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Custom Age Imputer**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From what I read on Kaggle, several imputing stategies were studied for missing Age values and a simple imputer with a median strategy should work just fine. For the sake of the exercise and a bit of practising with pandas, I will go for a custom imputer based on Title. Most of titles found for train set denote social or professional status and hence, according to me, contain some information about the age."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CustomAgeImputer(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, feature_name='Age', group_by_title=True, imp_func=np.nanmedian):\n",
    "        self.feature_name = feature_name\n",
    "        self.group_by_title = group_by_title\n",
    "        self.imp_func = imp_func\n",
    "    def fit(self, df, y=None):\n",
    "        self.df_median = df.groupby(['Title'])[self.feature_name].agg(self.imp_func)\n",
    "        self.med_M = self.imp_func(df.loc[df['Sex'] == 'male', 'Age'])\n",
    "        self.med_F = self.imp_func(df.loc[df['Sex'] == 'female', 'Age'])\n",
    "        return self\n",
    "    def transform(self, df, y=None):\n",
    "        if self.group_by_title:\n",
    "            for title in df['Title'].unique():\n",
    "                if title in self.df_median.dropna().index:\n",
    "                    df.loc[(df['Title'] == title) & (df[self.feature_name].isna()), self.feature_name] = self.df_median[title] \n",
    "                    #filna method with inplace not working here... should investigate what's wrong later for nicer code\n",
    "                else:\n",
    "                    df.loc[(df['Title'] == title) & (df['Sex'] == 'male') & (df[self.feature_name].isna()), self.feature_name] = self.med_M\n",
    "                    df.loc[(df['Title'] == title) & (df['Sex'] == 'female') & (df[self.feature_name].isna()), self.feature_name] = self.med_F\n",
    "        else:\n",
    "            df.loc[(df['Sex'] == 'male') & (df[self.feature_name].isna()), self.feature_name] = self.med_M\n",
    "            df.loc[(df['Sex'] == 'female') & (df[self.feature_name].isna()), self.feature_name] = self.med_F\n",
    "        return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Inspired from stackoverflow.com/questions/25239958 and copy from https://github.com/ageron/handson-ml/blob/master/03_classification.ipynb\n",
    "class MostFrequentImputer(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, df, y=None):\n",
    "        self.most_frequent_ = pd.Series([df[c].value_counts().index[0] for c in df],\n",
    "                                        index=df.columns)\n",
    "        return self\n",
    "    def transform(self, df, y=None):\n",
    "        return df.fillna(self.most_frequent_)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **DataFrame Selector**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's finally add a DataFrame Selector Class so we can easily select relevant features within pipelines later if needed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DataFrameSelector(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, feature_names):\n",
    "        self.feature_names = feature_names\n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    def transform(self, X):\n",
    "        return X[self.feature_names]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Pipelines**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "One last step before training models, we need to build the preprocessing pipelines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "add_imp_pipeline = Pipeline([('feat_add', FeatureAdder()),\n",
    "                             ('age_imp', CustomAgeImputer()),\n",
    "                             ('age_bucket_add', AgeBucketAdder())])\n",
    "\n",
    "num_features = ['Fare']\n",
    "cat_features = ['Pclass', 'Sex', 'Embarked', 'Title', 'Family', 'Age_bucket'] #not sure if Embarked is relevant. We'll see that later.\n",
    "feat_l = [[1, 2], ['female'], ['S', 'C'], ['Master', 'Miss'], ['single', 'sm_fam'], [0.0, 15.0, 30.0, 45.0]]\n",
    "\n",
    "num_pipeline = Pipeline([\n",
    "    ('num_imp', SimpleImputer(strategy='median')), #could have build imp based on pclass but no missing value for Fare in test set\n",
    "    ('scaler', StandardScaler())\n",
    "])\n",
    "\n",
    "cat_pipeline = Pipeline([\n",
    "    ('cat_imp', MostFrequentImputer()),\n",
    "    ('enc', OneHotEncoder(sparse=False, handle_unknown='ignore', categories=feat_l))\n",
    "])\n",
    "\n",
    "union_pipeline = ColumnTransformer([\n",
    "    ('num', num_pipeline, num_features),\n",
    "    ('cat', cat_pipeline, cat_features)\n",
    "])\n",
    "\n",
    "preproc_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline)\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's give a try and check that everything looks fine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files were previously downloaded, loading file from:  ./Data/\n"
     ]
    }
   ],
   "source": [
    "datasets = fetch_datasets()\n",
    "df_train = datasets['train.csv']\n",
    "df_prep = df_train.copy()\n",
    "df_test = datasets['test.csv']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_prep = preproc_pipeline.fit_transform(df_prep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(891, 14)\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>Fare</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>female</th>\n",
       "      <th>S</th>\n",
       "      <th>C</th>\n",
       "      <th>Master</th>\n",
       "      <th>Miss</th>\n",
       "      <th>single</th>\n",
       "      <th>sm_fam</th>\n",
       "      <th>0.0</th>\n",
       "      <th>15.0</th>\n",
       "      <th>30.0</th>\n",
       "      <th>45.0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.502445</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.786845</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.488854</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.420730</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.486337</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-0.478116</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.395814</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-0.224083</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>-0.424256</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>-0.042956</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>Fare</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>female</th>\n",
       "      <th>S</th>\n",
       "      <th>C</th>\n",
       "      <th>Master</th>\n",
       "      <th>Miss</th>\n",
       "      <th>single</th>\n",
       "      <th>sm_fam</th>\n",
       "      <th>0.0</th>\n",
       "      <th>15.0</th>\n",
       "      <th>30.0</th>\n",
       "      <th>45.0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.502445</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.786845</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.488854</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.420730</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.486337</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-0.478116</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.395814</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-0.224083</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>-0.424256</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>-0.042956</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "col_name = num_features + list(itertools.chain.from_iterable(feat_l))\n",
    "print(X_prep.shape)\n",
    "display(pd.DataFrame(X_prep, columns=col_name).head(10))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally.... Everything looks fine and it will be very simple to modify or even tune preprocessing parameters thanks to pipelines. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Training models**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Short-list of promising models**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = df_train['Survived']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
      "           weights='uniform')\n",
      "Mean Accury scores for knn : 0.7823921802292588\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "LinearSVC(C=1.0, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='hinge', max_iter=1000, multi_class='ovr',\n",
      "     penalty='l2', random_state=None, tol=0.0001, verbose=0)\n",
      "Mean Accury scores for lin_svm : 0.7934777550788786\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
      "  decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',\n",
      "  max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
      "  tol=0.001, verbose=False)\n",
      "Mean Accury scores for svm : 0.7844884235614572\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
      "          intercept_scaling=1, max_iter=100, multi_class='warn',\n",
      "          n_jobs=None, penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False)\n",
      "Mean Accury scores for logit : 0.8047398138690273\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=None,\n",
      "            max_features=None, max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, presort=False, random_state=42,\n",
      "            splitter='best')\n",
      "Mean Accury scores for dt : 0.7980101577573488\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators='warn', n_jobs=None,\n",
      "            oob_score=False, random_state=42, verbose=0, warm_start=False)\n",
      "Mean Accury scores for rf : 0.8081361366473725\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "MLPClassifier(activation='relu', alpha=0.0001, batch_size='auto', beta_1=0.9,\n",
      "       beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
      "       hidden_layer_sizes=(100,), learning_rate='constant',\n",
      "       learning_rate_init=0.001, max_iter=200, momentum=0.9,\n",
      "       n_iter_no_change=10, nesterovs_momentum=True, power_t=0.5,\n",
      "       random_state=42, shuffle=True, solver='adam', tol=0.0001,\n",
      "       validation_fraction=0.1, verbose=False, warm_start=False)\n",
      "Mean Accury scores for mlp : 0.8070241743275451\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "BernoulliNB(alpha=1.0, binarize=0.0, class_prior=None, fit_prior=True)\n",
      "Mean Accury scores for nb_clf : 0.783440301895358\n"
     ]
    }
   ],
   "source": [
    "algo_d = {'knn': KNeighborsClassifier(),\n",
    "          'lin_svm': LinearSVC(loss='hinge'),\n",
    "          'svm': SVC(gamma='auto'),\n",
    "          'logit': LogisticRegression(),\n",
    "          'dt': DecisionTreeClassifier(random_state=42),\n",
    "          'rf': RandomForestClassifier(random_state=42),\n",
    "          'mlp': MLPClassifier(random_state=42),\n",
    "          }\n",
    "scores_d = {}\n",
    "for key, value in algo_d.items():\n",
    "    full_pipeline = Pipeline([\n",
    "        ('prep', add_imp_pipeline),\n",
    "        ('union', union_pipeline),\n",
    "        (key, value)\n",
    "    ])\n",
    "    print(value)\n",
    "    scores = cross_val_score(full_pipeline, df_train, y_train, cv=10, n_jobs=-1)\n",
    "    scores_d[key] = scores\n",
    "    print('Mean Accury scores for {} : {}'.format(key, scores.mean()))\n",
    "    print('~-' * 30)\n",
    "\n",
    "nb_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('df_sel', DataFrameSelector(feature_names=cat_features)),\n",
    "    ('cat', cat_pipeline),\n",
    "    ('b_nb', BernoulliNB())\n",
    "])\n",
    "print(BernoulliNB())\n",
    "scores = cross_val_score(nb_pipeline, df_train, y_train, cv=10, n_jobs=-1)\n",
    "print('Mean Accury scores for {} : {}'.format('nb_clf', scores.mean()))\n",
    "scores_d['b_nb'] = scores\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Accuracy')"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_score = pd.DataFrame(scores_d).stack().reset_index()\n",
    "g = sns.violinplot(df_score['level_1'], df_score[0])\n",
    "g.set_xlabel('Models')\n",
    "g.set_ylabel('Accuracy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since binary class proportion for the target feature is rather balanced (~40 VS ~60), accuracy is not too bad as a metric for classifier evaluation. ROC curves, auc scores could have been another option.<br>\n",
    "However, 10 fold cross-val scores for accuracy may varie a lot among a single classifier. Thus the median or the violin-box distribution might be a intereresting complement for the cross-val score estimation in addition of average score. Inter-clf score distributions and medians remain roughly similar, and thus it is tricky to establish a short-list of most-promising models. <br> <br>\n",
    "Here the dataset is pretty small and I will go with all the previous classifiers except DT (random forest should perform at least just as well) and naive_bayes bernoulli."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Fine-tuning classifiers**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Random Forests*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 50 candidates, totalling 250 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   41.8s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  3.1min\n",
      "[Parallel(n_jobs=-1)]: Done 250 out of 250 | elapsed:  5.0min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise-deprecating',\n",
       "          estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...rs='warn', n_jobs=None,\n",
       "            oob_score=False, random_state=42, verbose=0, warm_start=False))]),\n",
       "          fit_params=None, iid='warn', n_iter=50, n_jobs=-1,\n",
       "          param_distributions={'rf__n_estimators': [1000], 'rf__max_features': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F77BDD8>, 'rf__max_depth': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F7A7470>},\n",
       "          pre_dispatch='2*n_jobs', random_state=42, refit=True,\n",
       "          return_train_score='warn', scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "        ('prep', add_imp_pipeline),\n",
    "        ('union', union_pipeline),\n",
    "        ('rf', RandomForestClassifier(random_state=42))\n",
    "])\n",
    "\n",
    "param_dist = {'rf__n_estimators': [1000], 'rf__max_features': uniform(0.2, 0.7), 'rf__max_depth': randint(2, 9)}\n",
    "\n",
    "R_cv = RandomizedSearchCV(full_pipeline, param_distributions=param_dist, scoring='accuracy', n_iter=50, \n",
    "                          n_jobs=-1, verbose=2, cv=5, random_state=42)\n",
    "R_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8383838383838383 {'rf__max_depth': 7, 'rf__max_features': 0.848661206055656, 'rf__n_estimators': 1000}\n"
     ]
    }
   ],
   "source": [
    "print(R_cv.best_score_, R_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_saver(model, saving_name, saving_path=MODEL_SAVING_PATH):\n",
    "    from sklearn.externals import joblib\n",
    "    os.makedirs(MODEL_SAVING_PATH, exist_ok=True)\n",
    "    joblib.dump(model, os.path.join(saving_path, saving_name))\n",
    "    print('Model saved in ', saving_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(R_cv.best_estimator_, 'rf_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's have a look to mean feature importances computed for Random Forests. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(0.41423963289229854, 'female'), (0.2570248623117338, 'Fare'), (0.08181752544240835, 2), (0.07263040572797998, 1), (0.03952706451353138, 'Master'), (0.034386937954674554, 'Miss'), (0.020229231347220416, 'S'), (0.018610233182737666, 0.0), (0.015741908469049666, 15.0), (0.014104521529181787, 'single'), (0.012850585114821917, 30.0), (0.009985026768412836, 'C'), (0.008852064745948816, 45.0), (0.0, 'sm_fam')]\n"
     ]
    }
   ],
   "source": [
    "feature_importances = R_cv.best_estimator_.steps[2][1].feature_importances_ #3rd step in pipeline, 2nd element for estimator\n",
    "print(sorted(list(zip(feature_importances, col_name)), reverse=True))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We should also give a try to some extra randomness with Extra-trees performing using random tresholds for features at splitting."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 50 candidates, totalling 250 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   40.3s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  2.9min\n",
      "[Parallel(n_jobs=-1)]: Done 250 out of 250 | elapsed:  4.8min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise-deprecating',\n",
       "          estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ors='warn', n_jobs=None,\n",
       "           oob_score=False, random_state=42, verbose=0, warm_start=False))]),\n",
       "          fit_params=None, iid='warn', n_iter=50, n_jobs=-1,\n",
       "          param_distributions={'ext_tr__n_estimators': [1000], 'ext_tr__max_features': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F3B9390>, 'ext_tr__max_depth': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F3B9D30>},\n",
       "          pre_dispatch='2*n_jobs', random_state=42, refit=True,\n",
       "          return_train_score='warn', scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "        ('prep', add_imp_pipeline),\n",
    "        ('union', union_pipeline),\n",
    "        ('ext_tr', ExtraTreesClassifier(random_state=42))\n",
    "])\n",
    "\n",
    "param_dist = {'ext_tr__n_estimators': [1000], 'ext_tr__max_features': uniform(0.2, 0.7), 'ext_tr__max_depth': randint(2, 10)}\n",
    "\n",
    "R_cv = RandomizedSearchCV(full_pipeline, param_distributions=param_dist, scoring='accuracy', n_iter=50, \n",
    "                          n_jobs=-1, verbose=2, cv=5, random_state=42)\n",
    "R_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8215488215488216 {'ext_tr__max_depth': 5, 'ext_tr__max_features': 0.856986896311025, 'ext_tr__n_estimators': 1000}\n"
     ]
    }
   ],
   "source": [
    "print(R_cv.best_score_, R_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(R_cv.best_estimator_, 'ext_tr_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's define a final feature selector for other classifier pipelines using ordered feature importances from Random Forests."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "def top_feat_k(l, k, rev=True): \n",
    "    l_zip = list(zip(l, range(len(l))))\n",
    "    l_zip.sort(reverse=rev)\n",
    "    l_ix = [l_zip[x][1] for x in range(len(l))]\n",
    "    return (l_ix[0:k])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "class FinalFeatureSelector(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, feat_import=feature_importances, k=len(feature_importances)):\n",
    "        self.k = k\n",
    "        self.feat_import = feat_import\n",
    "    def fit(self, X, y=None):\n",
    "        return(self)\n",
    "    def transform(self, X, y=None):\n",
    "        return X[:, top_feat_k(l=self.feat_import, k=self.k)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Knn*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 84 candidates, totalling 420 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   25.1s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  1.8min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  4.2min\n",
      "[Parallel(n_jobs=-1)]: Done 420 out of 420 | elapsed:  4.9min finished\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\model_selection\\_search.py:841: DeprecationWarning: The default of the `iid` parameter will change from True to False in version 0.22 and will be removed in 0.24. This will change numeric results when test-set sizes are unequal.\n",
      "  DeprecationWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ki',\n",
       "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
       "           weights='uniform'))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid={'knn__n_neighbors': range(3, 10), 'feat_sel__k': range(3, 15)},\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector(feat_import=feature_importances)),\n",
    "    ('knn', KNeighborsClassifier())\n",
    "])\n",
    "param = {'knn__n_neighbors': range(3, 10), 'feat_sel__k': range(3, len(feature_importances)+1)}\n",
    "G_cv = GridSearchCV(full_pipeline, param_grid=param, scoring='accuracy', n_jobs=-1, verbose=2, cv=5)\n",
    "G_cv.fit(df_train, y_train)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8249158249158249 {'feat_sel__k': 5, 'knn__n_neighbors': 8}\n"
     ]
    }
   ],
   "source": [
    "print(G_cv.best_score_, G_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(G_cv.best_estimator_, 'knn_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*RBF SVM*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 100 candidates, totalling 500 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   24.2s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  1.8min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  4.1min\n",
      "[Parallel(n_jobs=-1)]: Done 500 out of 500 | elapsed:  5.8min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise-deprecating',\n",
       "          estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...bf', max_iter=-1, probability=True, random_state=None,\n",
       "  shrinking=True, tol=0.001, verbose=False))]),\n",
       "          fit_params=None, iid='warn', n_iter=100, n_jobs=-1,\n",
       "          param_distributions={'svm__kernel': ['rbf'], 'svm__C': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F4E3D30>, 'svm__gamma': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F4E3C18>, 'feat_sel__k': range(4, 15)},\n",
       "          pre_dispatch='2*n_jobs', random_state=42, refit=True,\n",
       "          return_train_score='warn', scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector(feat_import=feature_importances)),\n",
    "    ('svm', SVC(probability=True))#prob slow down training but useful later\n",
    "])\n",
    "\n",
    "param_dist = {'svm__kernel': ['rbf'], 'svm__C': reciprocal(1, 1000), 'svm__gamma': reciprocal(0.001, 1),\n",
    "         'feat_sel__k': range(4, len(feature_importances)+1)}\n",
    "R_cv = RandomizedSearchCV(full_pipeline, param_distributions=param_dist, scoring='accuracy', n_jobs=-1, verbose=2, \n",
    "                          n_iter=100, cv=5, random_state=42)\n",
    "R_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8305274971941639 {'feat_sel__k': 8, 'svm__C': 63.5835885667625, 'svm__gamma': 0.13311216080736882, 'svm__kernel': 'rbf'}\n"
     ]
    }
   ],
   "source": [
    "print(R_cv.best_score_, R_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(R_cv.best_estimator_, 'rbf_svm_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Linear SVC*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 100 candidates, totalling 500 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   23.6s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  1.7min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  4.0min\n",
      "[Parallel(n_jobs=-1)]: Done 500 out of 500 | elapsed:  5.6min finished\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\model_selection\\_search.py:841: DeprecationWarning: The default of the `iid` parameter will change from True to False in version 0.22 and will be removed in 0.24. This will change numeric results when test-set sizes are unequal.\n",
      "  DeprecationWarning)\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\svm\\base.py:931: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.\n",
      "  \"the number of iterations.\", ConvergenceWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise-deprecating',\n",
       "          estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ax_iter=2000,\n",
       "     multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
       "     verbose=0))]),\n",
       "          fit_params=None, iid='warn', n_iter=100, n_jobs=-1,\n",
       "          param_distributions={'svm__penalty': ['l2'], 'svm__loss': ['hinge', 'squared_hinge'], 'svm__C': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C86F238B00>, 'feat_sel__k': range(4, 15)},\n",
       "          pre_dispatch='2*n_jobs', random_state=42, refit=True,\n",
       "          return_train_score='warn', scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector(feat_import=feature_importances)),\n",
    "    ('svm', LinearSVC(max_iter=2000))\n",
    "])\n",
    "\n",
    "param_dist = {'svm__penalty': ['l2'], 'svm__loss': ['hinge', 'squared_hinge'], 'svm__C': reciprocal(1, 1000),\n",
    "         'feat_sel__k': range(4, len(feature_importances)+1)}\n",
    "R_cv = RandomizedSearchCV(full_pipeline, param_distributions=param_dist, scoring='accuracy', n_jobs=-1, verbose=2,\n",
    "                          cv=5, n_iter=100, random_state=42)\n",
    "R_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8047138047138047 {'feat_sel__k': 10, 'svm__C': 145.48552070164158, 'svm__loss': 'squared_hinge', 'svm__penalty': 'l2'}\n"
     ]
    }
   ],
   "source": [
    "print(R_cv.best_score_, R_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(R_cv.best_estimator_, 'lin_svm_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Logit*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 68 candidates, totalling 340 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   23.2s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  1.7min\n",
      "[Parallel(n_jobs=-1)]: Done 340 out of 340 | elapsed:  3.7min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...penalty='l2', random_state=None, solver='warn',\n",
       "          tol=0.0001, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'logit__penalty': ['l2'], 'logit__C': [1, 10, 30, 100], 'feat_sel__k': range(4, 14)}, {'logit__penalty': ['l1'], 'logit__C': [1, 10, 30, 100], 'feat_sel__k': range(8, 15)}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector(feat_import=feature_importances)),\n",
    "    ('logit', LogisticRegression())\n",
    "])\n",
    "\n",
    "param = [{'logit__penalty': ['l2'], 'logit__C': [1, 10, 30, 100],\n",
    "         'feat_sel__k': range(4, len(feature_importances))},\n",
    "         {'logit__penalty': ['l1'], 'logit__C': [1, 10, 30, 100],\n",
    "         'feat_sel__k': range(8, len(feature_importances)+1)}] #Let lasso sparse the model\n",
    "         #Should try Elasticnet penalty with scikit-learn 0.21\n",
    "         \n",
    "G_cv = GridSearchCV(full_pipeline, param_grid=param, scoring='accuracy', n_jobs=-1, verbose=2, cv=5)\n",
    "G_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8170594837261503 {'feat_sel__k': 13, 'logit__C': 10, 'logit__penalty': 'l2'}\n"
     ]
    }
   ],
   "source": [
    "print(G_cv.best_score_, G_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(G_cv.best_estimator_, 'logit_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Multi layer perceptron*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 180 candidates, totalling 900 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   54.2s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  4.3min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  8.0min\n",
      "[Parallel(n_jobs=-1)]: Done 640 tasks      | elapsed: 12.8min\n",
      "[Parallel(n_jobs=-1)]: Done 900 out of 900 | elapsed: 17.6min finished\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\neural_network\\multilayer_perceptron.py:562: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (300) reached and the optimization hasn't converged yet.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...=True, solver='adam', tol=0.0001,\n",
       "       validation_fraction=0.1, verbose=False, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'mlp__solver': ['lbfgs'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [1e-05, 0.0001, 0.001], 'feat_sel__k': range(5, 9)}, {'mlp__solver': ['sgd'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [1e-05, 0.0001, 0.001], 'mlp__early_stopping': [True, False], 'feat_sel__k': range(5, 9)}, {'mlp__solver': ['adam'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [1e-05, 0.0001, 0.001], 'mlp__early_stopping': [True, False], 'feat_sel__k': range(5, 9)}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector(feat_import=feature_importances)),\n",
    "    ('mlp', MLPClassifier(random_state=42, max_iter=300))\n",
    "])\n",
    "\n",
    "param = [{'mlp__solver': ['lbfgs'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [0.00001, 0.0001, 0.001],\n",
    "          'feat_sel__k': range(5, 9)},\n",
    "         {'mlp__solver': ['sgd'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [0.00001, 0.0001, 0.001],\n",
    "          'mlp__early_stopping': [True, False], 'feat_sel__k': range(5, 9)},\n",
    "         {'mlp__solver': ['adam'], 'mlp__hidden_layer_sizes': [100, 300, 500], 'mlp__alpha': [0.00001, 0.0001, 0.001],\n",
    "          'mlp__early_stopping': [True, False], 'feat_sel__k': range(5, 9)}]\n",
    "         \n",
    "G_cv = GridSearchCV(full_pipeline, param_grid=param, scoring='accuracy', n_jobs=-1, verbose=2, cv=5)\n",
    "G_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8338945005611672 {'feat_sel__k': 7, 'mlp__alpha': 1e-05, 'mlp__early_stopping': False, 'mlp__hidden_layer_sizes': 300, 'mlp__solver': 'adam'}\n"
     ]
    }
   ],
   "source": [
    "print(G_cv.best_score_, G_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(G_cv.best_estimator_, 'mlp_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Ensemble methods**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Voting clf*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('knn', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ki',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
      "           weights='uniform'))])), ('svm', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...',\n",
      "  max_iter=-1, probability=True, random_state=None, shrinking=True,\n",
      "  tol=0.001, verbose=False))])), ('logit', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False))])), ('rf', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...rs='warn', n_jobs=None,\n",
      "            oob_score=False, random_state=42, verbose=0, warm_start=False))])), ('mlp', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...True, solver='lbfgs', tol=0.0001,\n",
      "       validation_fraction=0.1, verbose=False, warm_start=False))])), ('bnb', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...sparse=False))])), ('b_nb', BernoulliNB(alpha=1.0, binarize=0.0, class_prior=None, fit_prior=True))]))]\n"
     ]
    }
   ],
   "source": [
    "algo_d = {'knn': KNeighborsClassifier(),\n",
    "          'svm': SVC(gamma='auto', probability=True),\n",
    "          'logit': LogisticRegression(),\n",
    "          'rf': RandomForestClassifier(random_state=42),\n",
    "          'mlp': MLPClassifier(random_state=42, solver='lbfgs')\n",
    "          }\n",
    "\n",
    "model_pipe_d = {}\n",
    "for key, value in algo_d.items():\n",
    "    full_pipeline = Pipeline([\n",
    "        ('prep', add_imp_pipeline),\n",
    "        ('union', union_pipeline),\n",
    "        (key, value)\n",
    "    ])\n",
    "    model_pipe_d[key] = full_pipeline\n",
    "    \n",
    "nb_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('df_sel', DataFrameSelector(feature_names=cat_features)),\n",
    "    ('cat', cat_pipeline),\n",
    "    ('b_nb', BernoulliNB())\n",
    "])\n",
    "\n",
    "model_pipe_d['bnb'] = nb_pipeline\n",
    "\n",
    "estimators = [(key, value) for key, value in model_pipe_d.items()]\n",
    "print(estimators)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  10 out of  10 | elapsed:   29.5s finished\n"
     ]
    }
   ],
   "source": [
    "hard_voting_clf = VotingClassifier(estimators=estimators, voting='hard')\n",
    "h_scores = cross_val_score(hard_voting_clf, df_train, y_train, cv=10, n_jobs=-1, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  10 out of  10 | elapsed:   30.7s finished\n"
     ]
    }
   ],
   "source": [
    "soft_voting_clf = VotingClassifier(estimators=estimators, voting='soft')\n",
    "s_scores = cross_val_score(soft_voting_clf, df_train, y_train, cv=10, n_jobs=-1, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Accury scores for hard_voting_clf : 0.8170993644308251\n",
      "Mean Accury scores for soft_voting_clf : 0.8160015889229373\n"
     ]
    }
   ],
   "source": [
    "print('Mean Accury scores for {} : {}'.format('hard_voting_clf', h_scores.mean()))\n",
    "print('Mean Accury scores for {} : {}'.format('soft_voting_clf', s_scores.mean()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['./Models/knn_clf.pkl', './Models/logit_clf.pkl', './Models/mlp_clf.pkl', './Models/rf_clf.pkl', './Models/rbf_svm_clf.pkl']\n"
     ]
    }
   ],
   "source": [
    "trained_models_vot = [os.path.join(MODEL_SAVING_PATH, file_name)for file_name in \n",
    "                      ('knn_clf.pkl', 'logit_clf.pkl', 'mlp_clf.pkl', 'rf_clf.pkl', 'rbf_svm_clf.pkl')]\n",
    "print(trained_models_vot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " ('./Models/knn_clf.pkl', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ki',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=8, p=2,\n",
      "           weights='uniform'))]))\n",
      "\n",
      " ('./Models/logit_clf.pkl', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False))]))\n",
      "\n",
      " ('./Models/mlp_clf.pkl', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...=True, solver='adam', tol=0.0001,\n",
      "       validation_fraction=0.1, verbose=False, warm_start=False))]))\n",
      "\n",
      " ('./Models/rf_clf.pkl', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...tors=1000, n_jobs=None, oob_score=False,\n",
      "            random_state=42, verbose=0, warm_start=False))]))\n",
      "\n",
      " ('./Models/rbf_svm_clf.pkl', Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...bf', max_iter=-1, probability=True, random_state=None,\n",
      "  shrinking=True, tol=0.001, verbose=False))]))\n"
     ]
    }
   ],
   "source": [
    "trained_estimators = [(key, joblib.load(key)) for key in trained_models_vot]\n",
    "for estimator in trained_estimators:\n",
    "    print(\"\\n\", estimator)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  10 out of  10 | elapsed:   32.9s finished\n"
     ]
    }
   ],
   "source": [
    "hard_voting_clf = VotingClassifier(estimators=trained_estimators, voting='hard')\n",
    "h_scores = cross_val_score(hard_voting_clf, df_train, y_train, cv=10, n_jobs=-1, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Accury scores for hard_voting_clf : 0.8362010554988082\n"
     ]
    }
   ],
   "source": [
    "print('Mean Accury scores for {} : {}'.format('hard_voting_clf', h_scores.mean()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\neural_network\\multilayer_perceptron.py:562: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (300) reached and the optimization hasn't converged yet.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "hard_voting_clf.fit_transform(df_train, y_train)\n",
    "model_saver(hard_voting_clf, 'hard_voting_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Bagging/Pasting*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "See Random Forests & Extra-Trees"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Boosting: AdaBoost*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "ada_clf = AdaBoostClassifier(DecisionTreeClassifier(max_depth=1), n_estimators=200, algorithm=\"SAMME.R\", learning_rate=0.5)\n",
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('ada_clf', ada_clf)\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  10 out of  10 | elapsed:    5.6s finished\n"
     ]
    }
   ],
   "source": [
    "scores = cross_val_score(full_pipeline, df_train, y_train, cv=10, n_jobs=-1, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Accury scores for ada_clf (Decision stumps) : 0.8260381908977413\n"
     ]
    }
   ],
   "source": [
    "print('Mean Accury scores for {} : {}'.format('ada_clf (Decision stumps)', scores.mean()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 168 candidates, totalling 840 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   28.2s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  2.2min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  5.0min\n",
      "[Parallel(n_jobs=-1)]: Done 640 tasks      | elapsed:  8.9min\n",
      "[Parallel(n_jobs=-1)]: Done 840 out of 840 | elapsed: 11.7min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...m='SAMME.R', base_estimator=None,\n",
       "          learning_rate=1.0, n_estimators=50, random_state=None))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid={'ada_clf__base_estimator': [DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=1,\n",
       "            max_features=None, max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_we...ators': [100, 200, 300, 400], 'ada_clf__learning_rate': [0.5, 0.75, 1], 'feat_sel__k': range(8, 15)},\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='accuracy', verbose=2)"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ada_clf = AdaBoostClassifier(algorithm=\"SAMME.R\")\n",
    "dt_1 = DecisionTreeClassifier(max_depth=1)\n",
    "dt_2 = DecisionTreeClassifier(max_depth=2)\n",
    "\n",
    "\n",
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector()),\n",
    "    ('ada_clf', ada_clf)\n",
    "])\n",
    "\n",
    "param = {'ada_clf__base_estimator': [dt_1, dt_2], 'ada_clf__n_estimators': [100, 200, 300, 400], \n",
    "         'ada_clf__learning_rate': [0.5, 0.75, 1], 'feat_sel__k': range(8, len(feature_importances)+1)}\n",
    "\n",
    "G_cv = GridSearchCV(full_pipeline, param_grid=param, scoring='accuracy', n_jobs=-1, verbose=2, cv=5)\n",
    "G_cv.fit(df_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8282828282828283 {'ada_clf__base_estimator': DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=1,\n",
      "            max_features=None, max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, presort=False, random_state=None,\n",
      "            splitter='best'), 'ada_clf__learning_rate': 1, 'ada_clf__n_estimators': 400, 'feat_sel__k': 11}\n"
     ]
    }
   ],
   "source": [
    "print(G_cv.best_score_, G_cv.best_params_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(G_cv.best_estimator_, 'ada_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Boosting: Extreme Gradient Boosting*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 300 candidates, totalling 1500 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:   28.3s\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed:  4.6min\n",
      "[Parallel(n_jobs=-1)]: Done 640 tasks      | elapsed:  8.3min\n",
      "[Parallel(n_jobs=-1)]: Done 1005 tasks      | elapsed: 13.0min\n",
      "[Parallel(n_jobs=-1)]: Done 1450 tasks      | elapsed: 19.6min\n",
      "[Parallel(n_jobs=-1)]: Done 1500 out of 1500 | elapsed: 20.3min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "RandomizedSearchCV(cv=5, error_score='raise-deprecating',\n",
       "          estimator=Pipeline(memory=None,\n",
       "     steps=[('prep', Pipeline(memory=None,\n",
       "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
       "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...\n",
       "       reg_alpha=0, reg_lambda=1, scale_pos_weight=1, seed=None,\n",
       "       silent=True, subsample=1))]),\n",
       "          fit_params=None, iid='warn', n_iter=300, n_jobs=-1,\n",
       "          param_distributions={'xgb_clf__n_estimators': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C870CB7DA0>, 'xgb_clf__max_depth': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000001C870CB77F0>, 'xgb_clf__subsample': <scipy.stats._distn_infrastructure.rv_frozen object at 0x0...py.stats._distn_infrastructure.rv_frozen object at 0x000001C870CAB5C0>, 'feat_sel__k': range(8, 15)},\n",
       "          pre_dispatch='2*n_jobs', random_state=42, refit=True,\n",
       "          return_train_score='warn', scoring=None, verbose=2)"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb_clf = xgb.XGBClassifier()\n",
    "full_pipeline = Pipeline([\n",
    "    ('prep', add_imp_pipeline),\n",
    "    ('union', union_pipeline),\n",
    "    ('feat_sel', FinalFeatureSelector()),\n",
    "    ('xgb_clf', xgb_clf)\n",
    "])\n",
    "param_dist = {'xgb_clf__n_estimators': randint(100, 150), \"xgb_clf__max_depth\": randint(2, 7), 'xgb_clf__subsample': uniform(0.6, 0.3),\n",
    "              'xgb_clf__colsammple_bytree': uniform(0.7, 0.3), 'xgb_clf__gamma': expon(0, 5), 'xgb_clf__eta': uniform(0.03, 0.3),\n",
    "              'feat_sel__k': range(8, len(feature_importances)+1)}\n",
    "\n",
    "R_cv = RandomizedSearchCV(full_pipeline, param_distributions=param_dist, n_iter=300, cv=5, verbose=2, n_jobs=-1, random_state=42)\n",
    "R_cv.fit(df_train, y_train)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8439955106621774 {'feat_sel__k': 12, 'xgb_clf__colsammple_bytree': 0.8920008900510741, 'xgb_clf__eta': 0.097410353091804, 'xgb_clf__gamma': 0.9923183371506022, 'xgb_clf__max_depth': 4, 'xgb_clf__n_estimators': 104, 'xgb_clf__subsample': 0.6309926871372047}\n"
     ]
    }
   ],
   "source": [
    "print(R_cv.best_score_, R_cv.best_params_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's have a quick look at our hyper parameter search space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 936x432 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axarr = plt.subplots(2, 3, figsize=(13, 6))\n",
    "for i in range(2):\n",
    "    for j in range(3):\n",
    "        key = list(R_cv.best_params_.keys())[j + i*j +1] #+1 to skip feat_sel\n",
    "        best_val = R_cv.best_params_[key]\n",
    "        ax = axarr[i, j]\n",
    "        sns.distplot(R_cv.cv_results_['param_'+key].filled().astype(np.float32), ax=ax)\n",
    "        ax.set_title(re.split('__', key)[1])\n",
    "        ax.axvline(x=best_val, c=\"blue\")\n",
    "        ax.legend(['best', 'dist'])\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "model_saver(R_cv.best_estimator_, 'xgb_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Stacking*"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's try to build a singlelayer stacking ensemble"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "trained_models_path = [os.path.join(MODEL_SAVING_PATH, file_name)for file_name in \n",
    "                      ('knn_clf.pkl', 'logit_clf.pkl', 'mlp_clf.pkl', 'rf_clf.pkl', 'rbf_svm_clf.pkl')]\n",
    "tuned_estimators = [(key, joblib.load(key)) for key in trained_models_path]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ki',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=8, p=2,\n",
      "           weights='uniform'))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...=True, solver='adam', tol=0.0001,\n",
      "       validation_fraction=0.1, verbose=False, warm_start=False))])\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...tors=1000, n_jobs=None, oob_score=False,\n",
      "            random_state=42, verbose=0, warm_start=False))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...bf', max_iter=-1, probability=True, random_state=None,\n",
      "  shrinking=True, tol=0.001, verbose=False))])\n"
     ]
    }
   ],
   "source": [
    "first_layer_predictions = []\n",
    "for tuple in tuned_estimators:\n",
    "    estimator = tuple[1]\n",
    "    print(\"\\n\", estimator)\n",
    "    y_pred = cross_val_predict(estimator, df_train, y_train, method='predict_proba', n_jobs=-1, cv=5)\n",
    "    first_layer_predictions.append(y_pred[:, 1]) #cross val predict return proba for each class \n",
    "\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.        , 0.05585816, 0.10949213, 0.07344884, 0.17815104],\n",
       "       [1.        , 0.96781561, 0.98793834, 0.993     , 0.88581555],\n",
       "       [0.625     , 0.50862372, 0.33238464, 0.36735851, 0.1780791 ],\n",
       "       ...,\n",
       "       [0.625     , 0.47182458, 0.27934757, 0.31081835, 0.20095701],\n",
       "       [0.625     , 0.58732617, 0.31648037, 0.56822358, 0.20545482],\n",
       "       [0.        , 0.11756442, 0.1012727 , 0.05232193, 0.20507301]])"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "blender_train_set = np.array(first_layer_predictions).T\n",
    "display(blender_train_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=1000, n_jobs=-1,\n",
       "            oob_score=True, random_state=42, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_blender_clf = RandomForestClassifier(oob_score=True, n_jobs=-1, n_estimators=1000, random_state=42)\n",
    "rf_blender_clf.fit(blender_train_set, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.813692480359147\n"
     ]
    }
   ],
   "source": [
    "print(rf_blender_clf.oob_score_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_depths = [1, 2, 3]\n",
    "max_features = [3, 4, 5]\n",
    "cv_results = []\n",
    "for max_depth in max_depths:\n",
    "    for max_feature in max_features:\n",
    "        rf_blender_clf = RandomForestClassifier(oob_score=True, n_jobs=-1, n_estimators=1000, random_state=42,\n",
    "                                                max_features=max_feature, max_depth=max_depth)\n",
    "        rf_blender_clf.fit(blender_train_set, y_train)\n",
    "        cv_results.append({\"max_depth\" : max_depth, \"max_features\": max_feature, \"oob_score\" : rf_blender_clf.oob_score_})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "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>max_depth</th>\n",
       "      <th>max_features</th>\n",
       "      <th>oob_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.839506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.837262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>0.833895</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>0.831650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>0.831650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0.827160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>0.827160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.824916</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>0.824916</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<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>max_depth</th>\n",
       "      <th>max_features</th>\n",
       "      <th>oob_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0.839506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>0.837262</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>3</td>\n",
       "      <td>5</td>\n",
       "      <td>0.833895</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>0.831650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "      <td>0.831650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0.827160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>0.827160</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.824916</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>0.824916</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cv_df = pd.DataFrame(cv_results)\n",
    "cv_df.sort_values(by='oob_score', ascending=False, inplace=True)\n",
    "display(cv_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "rf_blender_clf = RandomForestClassifier(oob_score=True, n_jobs=-1, n_estimators=1000, random_state=42, max_depth=1, max_features=5)\n",
    "rf_blender_clf.fit(blender_train_set, y_train)\n",
    "model_saver(rf_blender_clf, 'rf_blender_clf.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(0.608, 'rf_clf'), (0.335, 'mlp_clf'), (0.032, 'rbf_svm_clf'), (0.02, 'knn_clf'), (0.005, 'logit_clf')]\n"
     ]
    }
   ],
   "source": [
    "feature_importances = rf_blender_clf.feature_importances_ \n",
    "print(sorted(list(zip(feature_importances, ['knn_clf', 'logit_clf', 'mlp_clf', 'rf_clf', 'rbf_svm_clf'])), reverse=True))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Predicting test set & preparing submissions**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = R_cv.predict(df_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_values = df_test.index.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "def Kaggle_sub(y_values, id_key='PassengerId', id_values=id_values, y_key='Survived', save_path='./Submissions', name='final_sub'):\n",
    "    df_sub = pd.DataFrame({id_key: id_values, y_key: y_values})\n",
    "    os.makedirs(save_path, exist_ok=True)\n",
    "    df_sub.to_csv(os.path.join(save_path, name+'.csv'), index=False)\n",
    "    print('{} was saved at {}'.format(name, save_path))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "xgb_clf was saved at ./Submissions\n"
     ]
    }
   ],
   "source": [
    "Kaggle_sub(y_values=y_pred, name='xgb_clf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_file_paths = [os.path.join(MODEL_SAVING_PATH, file) for file in os.listdir(MODEL_SAVING_PATH)]\n",
    "model_names = [re.search('/\\w*/(.*?)\\.pkl', file_path).group(1) for file_path in model_file_paths]\n",
    "model_file_paths.remove('./Models/rf_blender_clf.pkl')#need to prepare df_test prior to predictions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ada_clf was saved at ./Submissions\n",
      "ext_tr_clf was saved at ./Submissions\n",
      "hard_voting_clf was saved at ./Submissions\n",
      "knn_clf was saved at ./Submissions\n",
      "lin_svm_clf was saved at ./Submissions\n",
      "logit_clf was saved at ./Submissions\n",
      "mlp_clf was saved at ./Submissions\n",
      "rbf_svm_clf was saved at ./Submissions\n",
      "rf_blender_clf was saved at ./Submissions\n",
      "rf_clf was saved at ./Submissions\n"
     ]
    }
   ],
   "source": [
    "for i in range(len(model_file_paths)):\n",
    "    model = joblib.load(model_file_paths[i])\n",
    "    y_pred = model.predict(df_test)\n",
    "    Kaggle_sub(name=model_names[i], y_values=y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's finally get predictions on test set for the rf blender trained earlier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...ki',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=8, p=2,\n",
      "           weights='uniform'))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...=True, solver='adam', tol=0.0001,\n",
      "       validation_fraction=0.1, verbose=False, warm_start=False))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...tors=1000, n_jobs=None, oob_score=False,\n",
      "            random_state=42, verbose=0, warm_start=False))])\n",
      "\n",
      " Pipeline(memory=None,\n",
      "     steps=[('prep', Pipeline(memory=None,\n",
      "     steps=[('feat_add', FeatureAdder(add_family=True, relevant_nb=3)), ('age_imp', CustomAgeImputer(feature_name='Age', group_by_title=True,\n",
      "         imp_func=<function nanmedian at 0x000001C869560B70>)), ('age_bucket_add', AgeBucketAdder(bucket_cap=60, bucket_...bf', max_iter=-1, probability=True, random_state=None,\n",
      "  shrinking=True, tol=0.001, verbose=False))])\n"
     ]
    }
   ],
   "source": [
    "first_layer_predictions_test = []\n",
    "for tuple in tuned_estimators:\n",
    "    estimator = tuple[1]\n",
    "    print(\"\\n\", estimator)\n",
    "    y_pred = estimator.predict_proba(df_test)\n",
    "    first_layer_predictions_test.append(y_pred[:, 1]) #cross val predict return proba for each class \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.125     , 0.0973147 , 0.11533647, 0.09808368, 0.18084736],\n",
       "       [0.875     , 0.39042611, 0.58902918, 0.41912481, 0.31562791],\n",
       "       [0.        , 0.16138245, 0.14728242, 0.1249521 , 0.18077995],\n",
       "       ...,\n",
       "       [0.        , 0.0639328 , 0.11133759, 0.04425423, 0.18083837],\n",
       "       [0.25      , 0.0639633 , 0.10824612, 0.12299177, 0.18084209],\n",
       "       [0.5       , 0.59787291, 0.48131907, 0.3835006 , 0.42147026]])"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "blender_test_set = np.array(first_layer_predictions_test).T\n",
    "display(blender_test_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_blender_clf was saved at ./Submissions\n"
     ]
    }
   ],
   "source": [
    "y_pred = rf_blender_clf.predict(blender_test_set)\n",
    "Kaggle_sub(y_values=y_pred, name='rf_blender_clf')\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}