Домашнее_задание_3_(26_07_2024).ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Аттестация 2 + Домашнее задание № 3\n",
    "\n",
    "Порядок действий.\n",
    "\n",
    "1. Предобработка.\n",
    "2. Анализ данных:\n",
    "- Выполните первичный анализ данных: статистические характеристики, распределение признаков, корреляционный анализ.\n",
    "- Постройте визуализации для лучшего понимания структуры данных.\n",
    "3. Построение модели логистической регрессии:\n",
    "- Постройте модель логистической регрессии для бинарной классификации.\n",
    "- Оцените качество модели с использованием метрик точности (accuracy), полноты (recall), точности (precision) и F1-меры.\n",
    "- Постройте ROC-кривую и вычислите AUC.\n",
    "4. Построение модели дерева решений, оценка качества модели с использованием соответствующих метрик.\n",
    "5. Построение модели случайного леса, оценка качества модели с использованием соответствующих метрик.\n",
    "6. Сравнение результатов всех построенных моделей.\n",
    "\n",
    "Обсуждение преимуществ и недостатков каждой модели, а также их применимость к данному набору данных."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV, RepeatedStratifiedKFold\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score, roc_curve, classification_report\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from numpy import mean\n",
    "from catboost import CatBoostClassifier\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "data_1 = sns.load_dataset(\"titanic\")  # Загружаем датасет\n",
    "\n",
    "# Заполним пропуски в столбце \"age\" медианными значениями и сохраним данные в новом столбце \"age_median\"\n",
    "data_1['age_median'] = data_1['age'].fillna(data_1['age'].median())\n",
    "\n",
    "'''Исключаем из работы часть столбцов.\n",
    "Столбец \"embarked\" содержит первую букву из названия городов с портами, где сели пассажиры. Эти же данные уже есть\n",
    "в столбце \"embark_town\".\n",
    "Столбец \"class\" содержит категориальные данные, которые по сути, дублируются в столбеце \"pclass\", в уже нужном нам виде.\n",
    "Столбец \"deck\" (\"палуба\") исключаем, т.к. отсутствует большая часть данных и заполнить пробелы каким-либо способом\n",
    "не возможно.\n",
    "Столбец \"alive\", т.к. он по сути дублирует столбец \"survived\"'''\n",
    "\n",
    "data_1.drop(['age', 'embarked', 'class', 'deck', 'alive'], axis = 1 , inplace = True)\n",
    "\n",
    "# Заполняем две пустые ячейки в столбце \"embark_town\", значениями с предыдущих строк\n",
    "data_1['embark_town'] = data_1['embark_town'].ffill()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Время работы в секундах: 0.04188871383666992\n",
      "Accuracy: 0.8171641791044776\n",
      "Precision: 0.803921568627451\n",
      "Recall: 0.7387387387387387\n",
      "F1 Score: 0.7699530516431925\n",
      "ROC AUC: 0.8830550295518448\n"
     ]
    }
   ],
   "source": [
    "start_1 = time.time()  # Засекаем время работы кода\n",
    "\n",
    "# Переводим все объектные данные в дамми-переменные\n",
    "new_df = pd.get_dummies(data_1, columns = ['sex', 'who', 'embark_town'])\n",
    "\n",
    "# Масштабирование числовых признаков\n",
    "scaler = StandardScaler()\n",
    "new_df[['pclass', 'sibsp', 'parch', 'fare', 'age_median']] = scaler.fit_transform(new_df[['pclass', 'sibsp', 'parch', 'fare', 'age_median']])\n",
    "# new_df[['fare', 'age_median']] = scaler.fit_transform(new_df[['fare', 'age_median']])  # Можно и менее глобально масштабировать, но от этого скоры становятся чуть меньше\n",
    "\n",
    "# Отделение признаков от целевой переменной\n",
    "X = new_df.drop(columns = ['survived'])\n",
    "y = new_df['survived']\n",
    "\n",
    "# Разделение данных на тренировочную и тестовую выборки\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state = 42)\n",
    "\n",
    "# Обучение модели логистической регрессии\n",
    "model = LogisticRegression()\n",
    "model.fit(X_train, y_train)\n",
    "\n",
    "# Предсказания\n",
    "y_pred = model.predict(X_test)\n",
    "y_proba = model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "# Оценка модели\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f1 = f1_score(y_test, y_pred)\n",
    "roc_auc = roc_auc_score(y_test, y_proba)\n",
    "\n",
    "finish_1 = time.time()  # Останавливаем таймер и считаем время\n",
    "res_1 = finish_1 - start_1\n",
    "\n",
    "print('Время работы в секундах:', res_1)\n",
    "print(f'Accuracy: {accuracy}')\n",
    "print(f'Precision: {precision}')\n",
    "print(f'Recall: {recall}')\n",
    "print(f'F1 Score: {f1}')\n",
    "print(f'ROC AUC: {roc_auc}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Построение ROC-кривой\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_test, y_proba)\n",
    "plt.figure()\n",
    "plt.plot(fpr, tpr, color = 'darkorange', lw = 2, label = f'ROC curve (area = {roc_auc:.2f})')\n",
    "plt.plot([0, 1], [0, 1], color = 'navy', lw = 2, linestyle = '--')\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.legend(loc = \"lower right\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.832089552238806\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.88      0.86       157\n",
      "           1       0.82      0.77      0.79       111\n",
      "\n",
      "    accuracy                           0.83       268\n",
      "   macro avg       0.83      0.82      0.83       268\n",
      "weighted avg       0.83      0.83      0.83       268\n",
      "\n",
      "Mean cross-validation score: 0.8103508882053857\n",
      "Время работы в секундах: 0.04089045524597168\n"
     ]
    }
   ],
   "source": [
    "# 4. Построение модели дерева решений, оценка качества модели с использованием соответствующих метрик.\n",
    "\n",
    "start_2 = time.time()  # Засекаем время работы кода\n",
    "\n",
    "# Переводим все объектные данные в дамми-переменные\n",
    "new_df_decision_tree_class = pd.get_dummies(data_1, columns = ['sex', 'who', 'embark_town'], drop_first = True)\n",
    "\n",
    "# Разделение данных на признаки и целевую переменную\n",
    "X = new_df_decision_tree_class.drop(columns = ['survived'])\n",
    "y = new_df_decision_tree_class['survived']\n",
    "\n",
    "# Разделение данных на обучающую и тестовую выборки\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.3, random_state = 42)\n",
    "\n",
    "# Создание и обучение модели дерева решений\n",
    "dt_clf = DecisionTreeClassifier(max_depth = 4, max_features = 5, random_state = 42)\n",
    "dt_clf.fit(X_train, y_train)\n",
    "\n",
    "# Предсказание на тестовой выборке\n",
    "y_pred = dt_clf.predict(X_test)\n",
    "\n",
    "# Оценка модели на тестовой выборке\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "print(f'Accuracy: {accuracy}')\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Кросс-валидация\n",
    "scores = cross_val_score(dt_clf, X, y, cv = 5)\n",
    "print(f'Mean cross-validation score: {scores.mean()}')\n",
    "\n",
    "finish_2 = time.time()  # Останавливаем таймер и считаем время\n",
    "res_2 = finish_2 - start_2\n",
    "print('Время работы в секундах:', res_2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.8324022346368715\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.83      0.90      0.86       105\n",
      "           1       0.84      0.73      0.78        74\n",
      "\n",
      "    accuracy                           0.83       179\n",
      "   macro avg       0.83      0.82      0.82       179\n",
      "weighted avg       0.83      0.83      0.83       179\n",
      "\n",
      "Mean cross-validation score: 0.8395505617977529\n",
      "Время работы в секундах: 1.2601406574249268\n"
     ]
    }
   ],
   "source": [
    "# 5. Построение модели случайного леса, оценка качества модели с использованием соответствующих метрик.\n",
    "\n",
    "start_3 = time.time()  # Засекаем время работы кода\n",
    "\n",
    "# Переводим все объектные данные в дамми-переменные\n",
    "new_df_random_forest_class = pd.get_dummies(data_1, columns = ['sex', 'who', 'embark_town'], drop_first = True)\n",
    "\n",
    "# Разделение данных на признаки и целевую переменную\n",
    "X = new_df_decision_tree_class.drop(columns = ['survived'])\n",
    "y = new_df_decision_tree_class['survived']\n",
    "\n",
    "# Разделение данных на обучающую и тестовую выборки\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 42)\n",
    "\n",
    "# Создание и обучение модели дерева решений\n",
    "rf_clf = RandomForestClassifier(n_estimators = 50, max_depth = 7, max_features = 8, min_samples_leaf = 2, random_state = 42)\n",
    "\n",
    "# Определение значений 'max_depth' и 'max_features'\n",
    "# rf_clf_params = {'max_depth': range(2, 10), 'max_features': range(3, 10)}\n",
    "# forest_grid = GridSearchCV(rf_clf, rf_clf_params, cv = 3, n_jobs = -1, verbose = True)\n",
    "# forest_grid.fit(X_train, y_train)\n",
    "# forest_grid.best_params_, forest_grid.best_score_ # ({'max_depth': 7, 'max_features': 8}, 0.8394)\n",
    "\n",
    "rf_clf.fit(X_train, y_train)\n",
    "\n",
    "# Предсказание на тестовой выборке\n",
    "y_pred = rf_clf.predict(X_test)\n",
    "\n",
    "# Оценка модели на тестовой выборке\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "print(f'Accuracy: {accuracy}')\n",
    "print(classification_report(y_test, y_pred))\n",
    "\n",
    "# Кросс-валидация\n",
    "scores = cross_val_score(rf_clf, X, y, cv = 10)\n",
    "print(f'Mean cross-validation score: {scores.mean()}')\n",
    "\n",
    "finish_3 = time.time()  # Останавливаем таймер и считаем время\n",
    "res_3 = (finish_3 - start_3)\n",
    "print('Время работы в секундах:', res_3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "И так, подведём результаты основываясь на метрику Accuracy:\n",
    "\n",
    "LogisticRegression - 0.8171\n",
    "DecisionTreeClassifier - 0.8320\n",
    "RandomForestClassifier - 0.8324\n",
    "\n",
    "В данном случает, RandomForest показал наилучший результат, как в точности, так и Mean cross-validation, где отрыв от RandomForest более значительный - 0.8395 против 0.8103.\n",
    "Но, время работы RandomForest ощутимо дольше - 1.06 сек против 0.04 сек у DecisionTree. Время работы LogisticRegression примерно одинаковое с DecisionTree, но точность меньше.\n",
    "\n",
    "Можно сделать RandomForest быстрее, например, уменьшив параметр кросс-валидации до 3, тем самым сократив время работы до 4 сек, но потеряв в Mean cross-validation score. Выбор зависит от цели и ресурсов машины - требуются более высокие скоры или меньшие временные затраты на выполнение кода."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "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>Model</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "      <th>F1 Score</th>\n",
       "      <th>ROC AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CatBoost</td>\n",
       "      <td>0.815642</td>\n",
       "      <td>0.80597</td>\n",
       "      <td>0.72973</td>\n",
       "      <td>0.765957</td>\n",
       "      <td>0.884685</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      Model  Accuracy  Precision   Recall  F1 Score   ROC AUC\n",
       "0  CatBoost  0.815642    0.80597  0.72973  0.765957  0.884685"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Переводим все объектные данные в дамми-переменные\n",
    "new_df_cat_boost_class = pd.get_dummies(data_1, columns = ['sex', 'who', 'embark_town'], drop_first = True)\n",
    "\n",
    "# Разделение данных на признаки и целевую переменную\n",
    "X = new_df_cat_boost_class.drop(columns = ['survived'])\n",
    "y = new_df_cat_boost_class['survived']\n",
    "\n",
    "# Разделение данных на обучающую и тестовую выборки\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 42)\n",
    "\n",
    "# Обучение модели CatBoost и вывод метрик\n",
    "\n",
    "models = {'CatBoost': CatBoostClassifier(verbose = 0, max_depth = 7, learning_rate = 0.01, n_estimators = 100)}\n",
    "\n",
    "metrics = []\n",
    "\n",
    "for model_name, model in models.items():\n",
    "    model.fit(X_train, y_train)\n",
    "\n",
    "    y_pred = model.predict(X_test)\n",
    "    y_pred_proba = model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "    accuracy = accuracy_score(y_test, y_pred)\n",
    "    precision = precision_score(y_test, y_pred)\n",
    "    recall = recall_score(y_test, y_pred)\n",
    "    f1 = f1_score(y_test, y_pred)\n",
    "    roc_auc = roc_auc_score(y_test, y_pred_proba)\n",
    "\n",
    "    metrics.append({\n",
    "        'Model': model_name,\n",
    "        'Accuracy': accuracy,\n",
    "        'Precision': precision,\n",
    "        'Recall': recall,\n",
    "        'F1 Score': f1,\n",
    "        'ROC AUC': roc_auc\n",
    "    })\n",
    "\n",
    "metrics_df = pd.DataFrame(metrics)\n",
    "metrics_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.827180\n"
     ]
    }
   ],
   "source": [
    "# Оценка точности модели CatBoost, с применением метода RepeatedStratifiedKFold (результат стал лучше)\n",
    "\n",
    "model = CatBoostClassifier(verbose = 0, n_estimators = 50)\n",
    "cv = RepeatedStratifiedKFold(n_splits = 7, n_repeats = 4, random_state = 1)\n",
    "n_scores = cross_val_score(model, X, y, scoring = 'accuracy', cv = cv)\n",
    "print('Accuracy: %.6f' % (mean(n_scores)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'learning_rate': 0.001, 'max_depth': 7, 'n_estimators': 500}\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>Accuracy</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "      <th>F1 Score</th>\n",
       "      <th>ROC AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.854749</td>\n",
       "      <td>0.913793</td>\n",
       "      <td>0.716216</td>\n",
       "      <td>0.80303</td>\n",
       "      <td>0.925869</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Accuracy  Precision    Recall  F1 Score   ROC AUC\n",
       "0  0.854749   0.913793  0.716216   0.80303  0.925869"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Оценка метрик модели XGBoost с подбором гиперпараметров\n",
    "\n",
    "param_grid = {\n",
    "    'n_estimators': [100, 500],\n",
    "    'learning_rate': [0.001, 1],\n",
    "    'max_depth': [3, 7]\n",
    "}\n",
    "\n",
    "test_model = XGBClassifier()\n",
    "grid_search = GridSearchCV(estimator = test_model, param_grid = param_grid, cv = 3, scoring = 'accuracy')\n",
    "\n",
    "grid_search.fit(X_train, y_train)\n",
    "\n",
    "best_params = grid_search.best_params_\n",
    "print(f'Best Parameters: {best_params}')\n",
    "\n",
    "# Обучение модели с лучшими параметрами\n",
    "test_model.set_params(**best_params)\n",
    "test_model.fit(X, y)\n",
    "\n",
    "metrics = []\n",
    "y_pred = test_model.predict(X_test)\n",
    "y_pred_proba = test_model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "accuracy = accuracy_score(y_test, y_pred)\n",
    "precision = precision_score(y_test, y_pred)\n",
    "recall = recall_score(y_test, y_pred)\n",
    "f1 = f1_score(y_test, y_pred)\n",
    "roc_auc = roc_auc_score(y_test, y_pred_proba)\n",
    "\n",
    "metrics.append({\n",
    "    'Accuracy': accuracy,\n",
    "    'Precision': precision,\n",
    "    'Recall': recall,\n",
    "    'F1 Score': f1,\n",
    "    'ROC AUC': roc_auc\n",
    "})\n",
    "\n",
    "pd.DataFrame(metrics)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Модель XGBoost с подбором гиперпараметров показала наилучший результат, в сравнении с CatBoost с указании параметров пользователем."
   ]
  }
 ],
 "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": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}