Titanic-case-study.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "e96b56bd",
   "metadata": {},
   "source": [
    "Goal of this case study is to predict survivors based on Titanic dataset and compare different alghoritms with different hiperparameters. At the end we can choose the best one for this task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "979e8d01",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "df7e7f0f",
   "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>pclass</th>\n",
       "      <th>survived</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",
       "      <th>home.dest</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>Abbing, Mr. Anthony</td>\n",
       "      <td>male</td>\n",
       "      <td>42.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>C.A. 5547</td>\n",
       "      <td>7.55</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>Abbott, Master. Eugene Joseph</td>\n",
       "      <td>male</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>C.A. 2673</td>\n",
       "      <td>20.25</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "      <td>East Providence, RI</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>Abbott, Mr. Rossmore Edward</td>\n",
       "      <td>male</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>C.A. 2673</td>\n",
       "      <td>20.25</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "      <td>East Providence, RI</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>Abbott, Mrs. Stanton (Rosa Hunt)</td>\n",
       "      <td>female</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>C.A. 2673</td>\n",
       "      <td>20.25</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "      <td>East Providence, RI</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>Abelseth, Miss. Karen Marie</td>\n",
       "      <td>female</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>348125</td>\n",
       "      <td>7.65</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "      <td>Norway Los Angeles, CA</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   pclass  survived                              name     sex   age  sibsp  \\\n",
       "0       3         0               Abbing, Mr. Anthony    male  42.0      0   \n",
       "1       3         0     Abbott, Master. Eugene Joseph    male  13.0      0   \n",
       "2       3         0       Abbott, Mr. Rossmore Edward    male  16.0      1   \n",
       "3       3         1  Abbott, Mrs. Stanton (Rosa Hunt)  female  35.0      1   \n",
       "4       3         1       Abelseth, Miss. Karen Marie  female  16.0      0   \n",
       "\n",
       "   parch     ticket   fare cabin embarked               home.dest  \n",
       "0      0  C.A. 5547   7.55   NaN        S                     NaN  \n",
       "1      2  C.A. 2673  20.25   NaN        S     East Providence, RI  \n",
       "2      1  C.A. 2673  20.25   NaN        S     East Providence, RI  \n",
       "3      1  C.A. 2673  20.25   NaN        S     East Providence, RI  \n",
       "4      0     348125   7.65   NaN        S  Norway Los Angeles, CA  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('titanic.csv', \n",
    "                   decimal=',')\n",
    "data.head() "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "25321d3b",
   "metadata": {},
   "source": [
    "1. Data preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "9c2f0cae",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 1046 entries, 0 to 1045\n",
      "Data columns (total 12 columns):\n",
      " #   Column     Non-Null Count  Dtype  \n",
      "---  ------     --------------  -----  \n",
      " 0   pclass     1046 non-null   int64  \n",
      " 1   survived   1046 non-null   int64  \n",
      " 2   name       1046 non-null   object \n",
      " 3   sex        1046 non-null   object \n",
      " 4   age        1046 non-null   float64\n",
      " 5   sibsp      1046 non-null   int64  \n",
      " 6   parch      1046 non-null   int64  \n",
      " 7   ticket     1046 non-null   object \n",
      " 8   fare       1045 non-null   float64\n",
      " 9   cabin      272 non-null    object \n",
      " 10  embarked   1044 non-null   object \n",
      " 11  home.dest  685 non-null    object \n",
      "dtypes: float64(2), int64(4), object(6)\n",
      "memory usage: 98.2+ KB\n"
     ]
    }
   ],
   "source": [
    "data.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b81fa126",
   "metadata": {},
   "outputs": [],
   "source": [
    "y = data['survived'] #it's our variable that we'd like to predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2aa329a3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "732"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['ticket'].nunique() #it's something like ID of ticket, we can't reasonably \n",
    "#use this information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "8246b850",
   "metadata": {},
   "outputs": [],
   "source": [
    "data.drop(columns='ticket', axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f9117f15",
   "metadata": {},
   "outputs": [],
   "source": [
    "data.drop(columns='fare', axis=1, inplace=True)\n",
    "#First thing we can see is that fare - ticket price is object Dtype and should be float. \n",
    "#But I decided not to use this variable at all, beacouse: 1) we can't be sure about currency and\n",
    "#correctness (probably it's the summarized price for all people from the ticket) \n",
    "#AND this variable is higly correlated with cabin and pclass. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "7a56011c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "C    86\n",
       "B    63\n",
       "D    42\n",
       "E    38\n",
       "A    19\n",
       "F    18\n",
       "G     5\n",
       "T     1\n",
       "Name: cabin, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#There is big amount of Nan's BUT it could be very useful information so we'll try to do sth\n",
    "#We'll only use the letter sign - we assume that it's a deck designation. \n",
    "data.cabin.str[0].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "eddd6d11",
   "metadata": {},
   "outputs": [],
   "source": [
    "data[\"deck\"] = data.cabin.str[0].fillna(\"Unknown\").replace({\"G\":\"F\", \"T\":\"F\"})\n",
    "#There is very little\n",
    "#samples in classes T and G so we can add them to class F. Nan's we turn into class Unknown."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4b963a0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "data.drop(['cabin'], axis=1, inplace=True) #now we can delete column cabin"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "1d656c4c",
   "metadata": {},
   "outputs": [],
   "source": [
    "data['embarked'].fillna(data['embarked'].value_counts().index[0], inplace=True)\n",
    "#Column 'embarked' it's the port from which passengers departed. It seems as usefull information\n",
    "#but it's probably highly correlated with cabin (now deck) and it's probably more informative but \n",
    "#on the other hand it can be more informative for us than ID's of cabins that we've alreade \n",
    "#established it's hard to locate on ship\n",
    "#We'll use this column and fill Nan's with most common value which is 'S'. There is only 2 Nan's\n",
    "#so it shouldn't be a mistake."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "1ef40325",
   "metadata": {},
   "outputs": [],
   "source": [
    "#home.dest (home destination) is the column that \n",
    "#we will not use as it does not carry any information related to the disaster"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "e017253e",
   "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>pclass</th>\n",
       "      <th>age</th>\n",
       "      <th>sibsp</th>\n",
       "      <th>parch</th>\n",
       "      <th>sex_male</th>\n",
       "      <th>embarked_Q</th>\n",
       "      <th>embarked_S</th>\n",
       "      <th>deck_B</th>\n",
       "      <th>deck_C</th>\n",
       "      <th>deck_D</th>\n",
       "      <th>deck_E</th>\n",
       "      <th>deck_F</th>\n",
       "      <th>deck_Unknown</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>42.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3</td>\n",
       "      <td>13.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>16.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3</td>\n",
       "      <td>16.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   pclass   age  sibsp  parch  sex_male  embarked_Q  embarked_S  deck_B  \\\n",
       "0       3  42.0      0      0         1           0           1       0   \n",
       "1       3  13.0      0      2         1           0           1       0   \n",
       "2       3  16.0      1      1         1           0           1       0   \n",
       "3       3  35.0      1      1         0           0           1       0   \n",
       "4       3  16.0      0      0         0           0           1       0   \n",
       "\n",
       "   deck_C  deck_D  deck_E  deck_F  deck_Unknown  \n",
       "0       0       0       0       0             1  \n",
       "1       0       0       0       0             1  \n",
       "2       0       0       0       0             1  \n",
       "3       0       0       0       0             1  \n",
       "4       0       0       0       0             1  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = pd.get_dummies(data [['pclass', 'sex', 'age', 'sibsp', 'parch', \n",
    "                          'embarked', 'deck']], drop_first=True)\n",
    "X.head() #'sex','embarked' and 'deck' are categorical columns so they should be encoded."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bafcb5bc",
   "metadata": {},
   "source": [
    "2. Modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bfe461be",
   "metadata": {},
   "source": [
    "2.1 Decision Tree Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "7d7d5b9f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "03966d66",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=123)\n",
    "#now we split data into training and testing. We can assume basing on dataset size that 80:20 \n",
    "#should be correct ratio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5268d9a5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import KFold, StratifiedKFold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "99eb2ba3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7142857142857143\n"
     ]
    }
   ],
   "source": [
    "model_dt = DecisionTreeClassifier()\n",
    "model_dt.fit(X_train, y_train)\n",
    "y_pred_dt = model_dt.predict(X_test)\n",
    "acc_dt = accuracy_score(y_test, y_pred_dt)\n",
    "print(acc_dt)\n",
    "#~70% accuracy when we fit Decision Tree Classifier without parameter optimization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "7fb3d96f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'criterion': 'gini', 'max_depth': 12, 'min_samples_leaf': 5}\n",
      "0.7857142857142857\n"
     ]
    }
   ],
   "source": [
    "kfold = StratifiedKFold(10, shuffle=True, random_state=123)\n",
    "\n",
    "model_dt_opt = DecisionTreeClassifier()\n",
    "params = {\"min_samples_leaf\": np.arange(1, 21),\n",
    "          'max_depth': np.arange(1, 16),\n",
    "         'criterion': ['gini', 'entropy']}\n",
    "optimizer_dt = GridSearchCV(model_dt, \n",
    "                         params, \n",
    "                         scoring=\"accuracy\", \n",
    "                         cv=kfold, \n",
    "                         n_jobs=-1)\n",
    "optimizer_dt.fit(X_train, y_train)\n",
    "\n",
    "y_pred_dt = optimizer_dt.best_estimator_.predict(X_test)\n",
    "print(optimizer_dt.best_params_)\n",
    "print(accuracy_score(y_test, y_pred_dt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "2498bbaa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7904761904761904\n"
     ]
    }
   ],
   "source": [
    "model_dt_opt = DecisionTreeClassifier(criterion='gini', max_depth=None, min_samples_leaf=5)\n",
    "model_dt_opt.fit(X_train, y_train)\n",
    "y_pred_dt_opt = model_dt_opt.predict(X_test)\n",
    "acc_dt_opt = accuracy_score(y_test, y_pred_dt_opt)\n",
    "print(acc_dt_opt)\n",
    "#Decision Tree Classifier with optimal parameters. Accuracy increases to ~79%"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "117288d7",
   "metadata": {},
   "source": [
    "2.2 Logistic regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "f0dff24d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.pipeline import Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "ec71114f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7857142857142857\n"
     ]
    }
   ],
   "source": [
    "model_lr = LogisticRegression(max_iter=10e5)\n",
    "model_lr.fit(X_train, y_train)\n",
    "y_pred_lr = model_lr.predict(X_test)\n",
    "acc_lr = accuracy_score(y_test, y_pred_lr)\n",
    "print(acc_lr)\n",
    "#~78% accuracy when we use Linear Regression model without optimization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "17514805",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'model__C': 0.1, 'model__penalty': 'l2'}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Optimization\n",
    "kfold = StratifiedKFold(10, shuffle=True, random_state=123)\n",
    "\n",
    "pipeline = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', LogisticRegression(solver='saga'))\n",
    "])\n",
    "params = [\n",
    "    {\n",
    "        'model__penalty': ['none']\n",
    "    },\n",
    "    {\n",
    "        'model__penalty': ['l1', 'l2'],\n",
    "        'model__C': [0.01, 0.1, 1, 10, 100, 1000]\n",
    "    }\n",
    "]\n",
    "\n",
    "optimizer_lr = GridSearchCV(pipeline,\n",
    "                           params,\n",
    "                           scoring='accuracy',\n",
    "                           cv=kfold,\n",
    "                           n_jobs=-1)\n",
    "optimizer_lr.fit(X_train, y_train)\n",
    "optimizer_lr.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "093ba98f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7714285714285715\n"
     ]
    }
   ],
   "source": [
    "model_lr_opt = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', LogisticRegression(penalty='l2', C=0.1))\n",
    "])\n",
    "model_lr_opt.fit(X_train, y_train)\n",
    "y_pred_lr_opt = model_lr_opt.predict(X_test)\n",
    "acc_lr_opt = accuracy_score(y_test, y_pred_lr_opt)\n",
    "print(acc_lr_opt)\n",
    "#after optimization we are at the same level of accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "6c4f65a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "57aed9ae",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[96 21]\n",
      " [24 69]]\n"
     ]
    }
   ],
   "source": [
    "cm = metrics.confusion_matrix(y_test, y_pred_lr)\n",
    "print(cm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "729ed221",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'solver': 'lbfgs', 'best params': {'model__C': 0.1, 'model__penalty': 'l2'}}\n",
      "Accuracy: 0.7714285714285715\n",
      "{'solver': 'liblinear', 'best params': {'model__C': 1.0, 'model__penalty': 'l2'}}\n",
      "Accuracy: 0.7761904761904762\n",
      "{'solver': 'newton-cg', 'best params': {'model__C': 0.1, 'model__penalty': 'l2'}}\n",
      "Accuracy: 0.7714285714285715\n",
      "{'solver': 'sag', 'best params': {'model__C': 0.1, 'model__penalty': 'l2'}}\n",
      "Accuracy: 0.7714285714285715\n",
      "{'solver': 'saga', 'best params': {'model__C': 0.1, 'model__penalty': 'l2'}}\n",
      "Accuracy: 0.7714285714285715\n"
     ]
    }
   ],
   "source": [
    "#Optimization\n",
    "kfold = StratifiedKFold(10, shuffle=True, random_state=123)\n",
    "\n",
    "solvers_list = ['lbfgs', 'liblinear', 'newton-cg', 'sag', 'saga']\n",
    "C_params = [10e-2, 10e-1, 1, 10, 10e2, 10e3]\n",
    "\n",
    "for solver in solvers_list:\n",
    "    pipeline = Pipeline([\n",
    "        ('scaler', StandardScaler()),\n",
    "        ('model', LogisticRegression(solver=solver))\n",
    "    ])\n",
    "    if solver in ['lbfgs', 'newton-cg', 'sag']:\n",
    "        params = [\n",
    "            {\n",
    "                'model__penalty': ['none']\n",
    "            },\n",
    "            {\n",
    "                'model__penalty': ['l2'],\n",
    "                'model__C': C_params\n",
    "            }\n",
    "        ]\n",
    "        \n",
    "    elif solver == 'liblinear':\n",
    "            params = [\n",
    "            {\n",
    "                'model__penalty': ['l1', 'l2'],\n",
    "                'model__C': C_params\n",
    "            }\n",
    "        ]\n",
    "    elif solver == 'saga':\n",
    "            params = [\n",
    "            {\n",
    "                'model__penalty': ['none']\n",
    "            },\n",
    "            {\n",
    "                'model__penalty': ['l1', 'l2'],\n",
    "                'model__C': C_params\n",
    "            }\n",
    "        ]\n",
    "\n",
    "    optimizer_lr = GridSearchCV(pipeline,\n",
    "                                   params,\n",
    "                                   scoring='accuracy',\n",
    "                                   cv=kfold,\n",
    "                                   n_jobs=-1)\n",
    "    optimizer_lr.fit(X_train, y_train)\n",
    "    print({'solver': solver,\n",
    "           'best params': optimizer_lr.best_params_})\n",
    "    y_pred_lr = optimizer_lr.predict(X_test)\n",
    "    acc = accuracy_score(y_test, y_pred_lr)\n",
    "    print(f'Accuracy: {acc}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "e2a68fe3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7761904761904762\n"
     ]
    }
   ],
   "source": [
    "#as we can see the best logistic regression parameters for this task is solver liblinear with C parameter 1 and l2 model penalty\n",
    "#We can create new model with optimal parameters\n",
    "model_lr_opt = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', LogisticRegression(solver = 'liblinear', C=1, penalty = 'l2'))\n",
    "])\n",
    "\n",
    "model_lr_opt.fit(X_train, y_train)\n",
    "y_pred_lr_opt = model_lr_opt.predict(X_test)\n",
    "acc_lr_opt = accuracy_score(y_pred_lr_opt, y_test)\n",
    "print(acc_lr_opt)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "17c7c1e3",
   "metadata": {},
   "source": [
    "2.3 SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "fc72fac2",
   "metadata": {},
   "outputs": [],
   "source": [
    "#SVC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "c5fea028",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.svm import SVC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "db72c19c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.780952380952381\n"
     ]
    }
   ],
   "source": [
    "model_SVC = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', SVC(max_iter=10e5))\n",
    "])\n",
    "model_SVC.fit(X_train, y_train)\n",
    "y_pred_SVC = model_SVC.predict(X_test)\n",
    "acc_SVC = accuracy_score(y_test, y_pred_SVC)\n",
    "print(acc_SVC)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "762d4bf9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7952380952380952"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#SVC optimization\n",
    "\n",
    "model_SVC_opt = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', SVC(max_iter=10e7))\n",
    "])\n",
    "params = {'model__C':[0.01, 0.1, 1, 10],\n",
    "          'model__gamma':[1,0.1,0.001,0.0001], \n",
    "          'model__kernel':['linear','rbf']}\n",
    "\n",
    "optimizer_SVC = GridSearchCV(model_SVC_opt,\n",
    "                             params,\n",
    "                             scoring='accuracy',\n",
    "                             cv=kfold,\n",
    "                             n_jobs=-1\n",
    ")\n",
    "optimizer_SVC.fit(X_train, y_train)    \n",
    "y_pred_SVC_opt = optimizer_SVC.predict(X_test)\n",
    "accuracy_score(y_test, y_pred_SVC_opt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "93d19a30",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'model__C': 1, 'model__gamma': 0.1, 'model__kernel': 'rbf'}"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "optimizer_SVC.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "b4aa546a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.780952380952381\n"
     ]
    }
   ],
   "source": [
    "pipeline_SVC_opt = Pipeline([\n",
    "    ('scaler', StandardScaler()),\n",
    "    ('model', SVC(max_iter=10e7, C=1, gamma=0.1, kernel='rbf'))\n",
    "])\n",
    "model_SVC_opt.fit(X_train, y_train)\n",
    "y_SVC_opt = model_SVC_opt.predict(X_test)\n",
    "acc_SVC_opt = accuracy_score(y_test, y_SVC_opt)\n",
    "print(acc_SVC_opt)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e1cd99a6",
   "metadata": {},
   "source": [
    "2.4 Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b6ebcf24",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "144299e1",
   "metadata": {},
   "outputs": [],
   "source": [
    "# We'll use Random Forest Classifier with optimal parameters found in point 2.1.\n",
    "#The problem is to find optimal number of estimators."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "ad832afd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7904761904761904"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "{'criterion': 'gini', 'max_depth': 12, 'min_samples_leaf': 5}\n",
    "0.7904761904761904"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "57c2c762",
   "metadata": {},
   "outputs": [],
   "source": [
    "N_estimators = np.arange(3, 802 , 25)\n",
    "accuracies = []\n",
    "for n_trees in N_estimators:\n",
    "    model_rfc = RandomForestClassifier(n_trees, criterion='gini', \n",
    "                                       max_depth=12, min_samples_leaf=5)\n",
    "    model_rfc.fit(X_train, y_train)\n",
    "    y_pred_rfc = model_rfc.predict(X_test)\n",
    "    accuracies.append(accuracy_score(y_test, y_pred_rfc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "3bce60be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.style.use('default')\n",
    "plt.plot(N_estimators, accuracies)\n",
    "plt.grid()\n",
    "plt.show()\n",
    "#We can see some peaks but it's hard to interpret what number of trees is the \n",
    "#best, but probably there is no sense to test wider range than 800."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "c762fb8a",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Let's find parameters by GridSearch\n",
    "model_rfc_opt = RandomForestClassifier(n_trees, criterion='gini', \n",
    "                                       max_depth=12, min_samples_leaf=5)\n",
    "params = {'n_estimators': np.arange(100, 801, 25)}\n",
    "optimizer_rfc = GridSearchCV(model_rfc_opt,\n",
    "                            params,\n",
    "                            scoring='accuracy',\n",
    "                            cv=kfold,\n",
    "                            n_jobs=-1)\n",
    "optimizer_rfc.fit(X_train, y_train)\n",
    "y_pred_rfc = optimizer_rfc.best_estimator_.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "380392bb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'n_estimators': 200}"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "optimizer_rfc.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "bc34d96a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8333333333333334\n"
     ]
    }
   ],
   "source": [
    "#Let's create model with optimal parameters\n",
    "model_rfc_opt = RandomForestClassifier(200, criterion='gini', \n",
    "                                       max_depth=12, min_samples_leaf=5)\n",
    "model_rfc_opt.fit(X_train, y_train)\n",
    "y_pred_rfc = model_rfc_opt.predict(X_test)\n",
    "acc_rfc_opt = accuracy_score(y_test, y_pred_rfc)\n",
    "print(acc_rfc_opt)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d640631a",
   "metadata": {},
   "source": [
    "2.5 Bagging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "5968b742",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now we will test Bagging with SVM estimator and check accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "01c338ea",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import BaggingClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "e8e3a9c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "dec_tree = DecisionTreeClassifier(criterion='gini', max_depth=12, min_samples_leaf=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "27aa59b7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7857142857142857\n"
     ]
    }
   ],
   "source": [
    "model_bagging = BaggingClassifier(dec_tree)\n",
    "model_bagging.fit(X_train, y_train)\n",
    "y_pred_bagging = model_bagging.predict(X_test)\n",
    "acc_bagging = accuracy_score(y_test, y_pred_bagging)\n",
    "print(acc_bagging)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "5f13a3ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Now let's find optimal parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "b2107d04",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_bagging_opt = BaggingClassifier(dec_tree)\n",
    "params = {'n_estimators': np.arange(100, 501, 100),\n",
    "         'max_samples': [0.25, 0.5, 0.75, 1],\n",
    "         'max_features': [0.25, 0.5, 0.75, 1]}\n",
    "\n",
    "optimizer_bagging = GridSearchCV(model_bagging_opt,\n",
    "                                params,\n",
    "                                scoring='accuracy',\n",
    "                                cv=kfold,\n",
    "                                n_jobs=-1)\n",
    "optimizer_bagging.fit(X_train, y_train)\n",
    "y_pred_bagging_opt = optimizer_bagging.predict(X_test)\n",
    "accuracy_score(y_test, y_pred_bagging_opt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "310bef90",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'max_features': 0.75, 'max_samples': 0.75, 'n_estimators': 200}"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "optimizer_bagging.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "38458502",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7857142857142857\n"
     ]
    }
   ],
   "source": [
    "model_bagging_opt = BaggingClassifier(base_estimator=pipeline_SVC_opt, n_estimators=200, \n",
    "                                      max_samples=0.75, max_features=0.75)\n",
    "model_bagging_opt.fit(X_train, y_train)\n",
    "y_pred_bagging_opt = model_bagging.predict(X_test)\n",
    "acc_bagging_opt = accuracy_score(y_test, y_pred_bagging_opt)\n",
    "print(acc_bagging_opt)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1e3470de",
   "metadata": {},
   "source": [
    "Models summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "8961cfc2",
   "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>Models</th>\n",
       "      <th>Optimization</th>\n",
       "      <th>Accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>RandomForest</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0.833333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Decision Tree</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0.790476</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Logistic Regression</td>\n",
       "      <td>No</td>\n",
       "      <td>0.785714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Bagging</td>\n",
       "      <td>No</td>\n",
       "      <td>0.785714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Bagging</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0.785714</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>SVC</td>\n",
       "      <td>No</td>\n",
       "      <td>0.780952</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>SVC</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0.780952</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Logistic Regression</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0.776190</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>Decision Tree</td>\n",
       "      <td>No</td>\n",
       "      <td>0.714286</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                Models Optimization  Accuracy\n",
       "0         RandomForest          Yes  0.833333\n",
       "1        Decision Tree          Yes  0.790476\n",
       "2  Logistic Regression           No  0.785714\n",
       "3              Bagging           No  0.785714\n",
       "4              Bagging          Yes  0.785714\n",
       "5                  SVC           No  0.780952\n",
       "6                  SVC          Yes  0.780952\n",
       "7  Logistic Regression          Yes  0.776190\n",
       "8        Decision Tree           No  0.714286"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "models = ['Logistic Regression', 'Logistic Regression', 'Decision Tree', 'Decision Tree',\n",
    "         'RandomForest', 'SVC', 'SVC', 'Bagging', 'Bagging']\n",
    "cond = ['No', 'Yes', 'No', 'Yes', 'Yes', 'No', 'Yes', 'No', 'Yes']\n",
    "accuracies = [acc_lr, acc_lr_opt, acc_dt, acc_dt_opt, acc_rfc_opt, acc_SVC, acc_SVC_opt,\n",
    "            acc_bagging, acc_bagging_opt]\n",
    "models_sum = pd.DataFrame({'Models': models, 'Optimization': cond, 'Accuracy': accuracies})\n",
    "models_sum = models_sum.sort_values(by='Accuracy', ascending=False).reset_index(drop=True)\n",
    "models_sum"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "7f214a93",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "models_sum.pivot(index='Models', columns='Optimization', \n",
    "                 values='Accuracy').plot(kind='bar', rot=0)\n",
    "plt.ylim([0.7, 0.85])\n",
    "plt.xticks(rotation=45)\n",
    "plt.ylabel('Accuracy')\n",
    "plt.grid(axis='y', linestyle='--', alpha=0.5);"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a1ebb03",
   "metadata": {},
   "source": [
    "Conslusions: Among all the tested models RandomForest turned out to be the best (accuracy ~ 83,3%). Optimal performance of the alghortim was obtained with 200 trees, Gini criterion, maximum depth 12 and the minimum number of samples to split a node 5.There is significant difference in the non-optimized and optimized DecisionTree alghoritm, while in the others there are no big differences.\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}