1031 lines
115 KiB
Plaintext
1031 lines
115 KiB
Plaintext
|
{
|
|||
|
|
"cells": [
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"# Данные по инсультам\n",
|
|||
|
|
"\n",
|
|||
|
|
"Выведем информацию о столбцах датасета:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 441,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Index(['id', 'gender', 'age', 'hypertension', 'heart_disease', 'ever_married',\n",
|
|||
|
|
" 'work_type', 'Residence_type', 'avg_glucose_level', 'bmi',\n",
|
|||
|
|
" 'smoking_status', 'stroke'],\n",
|
|||
|
|
" dtype='object')\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>id</th>\n",
|
|||
|
|
" <th>gender</th>\n",
|
|||
|
|
" <th>age</th>\n",
|
|||
|
|
" <th>hypertension</th>\n",
|
|||
|
|
" <th>heart_disease</th>\n",
|
|||
|
|
" <th>ever_married</th>\n",
|
|||
|
|
" <th>work_type</th>\n",
|
|||
|
|
" <th>Residence_type</th>\n",
|
|||
|
|
" <th>avg_glucose_level</th>\n",
|
|||
|
|
" <th>bmi</th>\n",
|
|||
|
|
" <th>smoking_status</th>\n",
|
|||
|
|
" <th>stroke</th>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" </thead>\n",
|
|||
|
|
" <tbody>\n",
|
|||
|
|
" <tr>\n",
|
|||
|
|
" <th>0</th>\n",
|
|||
|
|
" <td>9046</td>\n",
|
|||
|
|
" <td>Male</td>\n",
|
|||
|
|
" <td>67.0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" <td>Yes</td>\n",
|
|||
|
|
" <td>Private</td>\n",
|
|||
|
|
" <td>Urban</td>\n",
|
|||
|
|
" <td>228.69</td>\n",
|
|||
|
|
" <td>36.6</td>\n",
|
|||
|
|
" <td>formerly smoked</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" <tr>\n",
|
|||
|
|
" <th>1</th>\n",
|
|||
|
|
" <td>51676</td>\n",
|
|||
|
|
" <td>Female</td>\n",
|
|||
|
|
" <td>61.0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>Yes</td>\n",
|
|||
|
|
" <td>Self-employed</td>\n",
|
|||
|
|
" <td>Rural</td>\n",
|
|||
|
|
" <td>202.21</td>\n",
|
|||
|
|
" <td>NaN</td>\n",
|
|||
|
|
" <td>never smoked</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" <tr>\n",
|
|||
|
|
" <th>2</th>\n",
|
|||
|
|
" <td>31112</td>\n",
|
|||
|
|
" <td>Male</td>\n",
|
|||
|
|
" <td>80.0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" <td>Yes</td>\n",
|
|||
|
|
" <td>Private</td>\n",
|
|||
|
|
" <td>Rural</td>\n",
|
|||
|
|
" <td>105.92</td>\n",
|
|||
|
|
" <td>32.5</td>\n",
|
|||
|
|
" <td>never smoked</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" <tr>\n",
|
|||
|
|
" <th>3</th>\n",
|
|||
|
|
" <td>60182</td>\n",
|
|||
|
|
" <td>Female</td>\n",
|
|||
|
|
" <td>49.0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>Yes</td>\n",
|
|||
|
|
" <td>Private</td>\n",
|
|||
|
|
" <td>Urban</td>\n",
|
|||
|
|
" <td>171.23</td>\n",
|
|||
|
|
" <td>34.4</td>\n",
|
|||
|
|
" <td>smokes</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" <tr>\n",
|
|||
|
|
" <th>4</th>\n",
|
|||
|
|
" <td>1665</td>\n",
|
|||
|
|
" <td>Female</td>\n",
|
|||
|
|
" <td>79.0</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" <td>0</td>\n",
|
|||
|
|
" <td>Yes</td>\n",
|
|||
|
|
" <td>Self-employed</td>\n",
|
|||
|
|
" <td>Rural</td>\n",
|
|||
|
|
" <td>174.12</td>\n",
|
|||
|
|
" <td>24.0</td>\n",
|
|||
|
|
" <td>never smoked</td>\n",
|
|||
|
|
" <td>1</td>\n",
|
|||
|
|
" </tr>\n",
|
|||
|
|
" </tbody>\n",
|
|||
|
|
"</table>\n",
|
|||
|
|
"</div>"
|
|||
|
|
],
|
|||
|
|
"text/plain": [
|
|||
|
|
" id gender age hypertension heart_disease ever_married \\\n",
|
|||
|
|
"0 9046 Male 67.0 0 1 Yes \n",
|
|||
|
|
"1 51676 Female 61.0 0 0 Yes \n",
|
|||
|
|
"2 31112 Male 80.0 0 1 Yes \n",
|
|||
|
|
"3 60182 Female 49.0 0 0 Yes \n",
|
|||
|
|
"4 1665 Female 79.0 1 0 Yes \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type Residence_type avg_glucose_level bmi smoking_status \\\n",
|
|||
|
|
"0 Private Urban 228.69 36.6 formerly smoked \n",
|
|||
|
|
"1 Self-employed Rural 202.21 NaN never smoked \n",
|
|||
|
|
"2 Private Rural 105.92 32.5 never smoked \n",
|
|||
|
|
"3 Private Urban 171.23 34.4 smokes \n",
|
|||
|
|
"4 Self-employed Rural 174.12 24.0 never smoked \n",
|
|||
|
|
"\n",
|
|||
|
|
" stroke \n",
|
|||
|
|
"0 1 \n",
|
|||
|
|
"1 1 \n",
|
|||
|
|
"2 1 \n",
|
|||
|
|
"3 1 \n",
|
|||
|
|
"4 1 "
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 441,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"import pandas as pd\n",
|
|||
|
|
"import matplotlib.pyplot as plt\n",
|
|||
|
|
"import seaborn as sns\n",
|
|||
|
|
"from sklearn.model_selection import train_test_split\n",
|
|||
|
|
"from imblearn.over_sampling import RandomOverSampler\n",
|
|||
|
|
"from sklearn.preprocessing import StandardScaler\n",
|
|||
|
|
"import featuretools as ft\n",
|
|||
|
|
"from sklearn.linear_model import LinearRegression\n",
|
|||
|
|
"from sklearn.model_selection import cross_val_score\n",
|
|||
|
|
"import time\n",
|
|||
|
|
"from sklearn.metrics import root_mean_squared_error, r2_score, mean_absolute_error\n",
|
|||
|
|
"from sklearn.ensemble import RandomForestRegressor\n",
|
|||
|
|
"\n",
|
|||
|
|
"df = pd.read_csv(\"..//..//static//csv//healthcare-dataset-stroke-data.csv\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(df.columns)\n",
|
|||
|
|
"df.head()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"### Определим бизнес цели и цели технического проекта.\n",
|
|||
|
|
"\n",
|
|||
|
|
"1. Улучшение диагностики и профилактики инсульта.\n",
|
|||
|
|
" * Бизнес-цель: повышение точности прогнозирования риска инсульта среди пациентов для более раннего лечебного вмешательства. Определение основных факторов риска для более целенаправленного подхода в медицинском обслуживании.\n",
|
|||
|
|
" * Цель технического проекта: разработка статистической модели, которая решает задачу классификации и предсказывает возможность возникновения инсульта у пациентов на основе имеющихся данных (возраст, гипертония, заболевания сердца и пр.), с целью выявления групп риска. Внедрение этой модели в систему поддержки принятия медицинских решений для врачей.\n",
|
|||
|
|
"2. Снижение расходов на лечение инсультов.\n",
|
|||
|
|
" * Бизнес-цель: снижение затрат на лечение инсульта путем более эффективного распределения медицинских ресурсов и направленных профилактических мер.\n",
|
|||
|
|
" * Цель технического проекта: создание системы оценки индивидуального риска инсульта для пациентов, что позволит медучреждениям проводить профилактические меры среди целевых групп, сокращая расходы на лечение.\n",
|
|||
|
|
"\n",
|
|||
|
|
"### И теперь проверим датасет на пустые значения:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 442,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"id 0\n",
|
|||
|
|
"gender 0\n",
|
|||
|
|
"age 0\n",
|
|||
|
|
"hypertension 0\n",
|
|||
|
|
"heart_disease 0\n",
|
|||
|
|
"ever_married 0\n",
|
|||
|
|
"work_type 0\n",
|
|||
|
|
"Residence_type 0\n",
|
|||
|
|
"avg_glucose_level 0\n",
|
|||
|
|
"bmi 201\n",
|
|||
|
|
"smoking_status 0\n",
|
|||
|
|
"stroke 0\n",
|
|||
|
|
"dtype: int64\n",
|
|||
|
|
"\n",
|
|||
|
|
"id False\n",
|
|||
|
|
"gender False\n",
|
|||
|
|
"age False\n",
|
|||
|
|
"hypertension False\n",
|
|||
|
|
"heart_disease False\n",
|
|||
|
|
"ever_married False\n",
|
|||
|
|
"work_type False\n",
|
|||
|
|
"Residence_type False\n",
|
|||
|
|
"avg_glucose_level False\n",
|
|||
|
|
"bmi True\n",
|
|||
|
|
"smoking_status False\n",
|
|||
|
|
"stroke False\n",
|
|||
|
|
"dtype: bool\n",
|
|||
|
|
"\n",
|
|||
|
|
"bmi процент пустых значений: %3.93\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Количество пустых значений признаков\n",
|
|||
|
|
"print(df.isnull().sum())\n",
|
|||
|
|
"\n",
|
|||
|
|
"print()\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Есть ли пустые значения признаков\n",
|
|||
|
|
"print(df.isnull().any())\n",
|
|||
|
|
"\n",
|
|||
|
|
"print()\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Процент пустых значений признаков\n",
|
|||
|
|
"for i in df.columns:\n",
|
|||
|
|
" null_rate = df[i].isnull().sum() / len(df) * 100\n",
|
|||
|
|
" if null_rate > 0:\n",
|
|||
|
|
" print(f\"{i} процент пустых значений: %{null_rate:.2f}\")"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"В столбце bmi можно заметить пустые значение. Заменим их на медиану:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 443,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"\n",
|
|||
|
|
"Количество пустых значений в каждом столбце после замены:\n",
|
|||
|
|
"id 0\n",
|
|||
|
|
"gender 0\n",
|
|||
|
|
"age 0\n",
|
|||
|
|
"hypertension 0\n",
|
|||
|
|
"heart_disease 0\n",
|
|||
|
|
"ever_married 0\n",
|
|||
|
|
"work_type 0\n",
|
|||
|
|
"Residence_type 0\n",
|
|||
|
|
"avg_glucose_level 0\n",
|
|||
|
|
"bmi 0\n",
|
|||
|
|
"smoking_status 0\n",
|
|||
|
|
"stroke 0\n",
|
|||
|
|
"dtype: int64\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Замена значений\n",
|
|||
|
|
"df[\"bmi\"] = df[\"bmi\"].fillna(df[\"bmi\"].median())\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Проверка на пропущенные значения после замены\n",
|
|||
|
|
"missing_values_after_drop = df.isnull().sum()\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Вывод результатов после замены\n",
|
|||
|
|
"print(\"\\nКоличество пустых значений в каждом столбце после замены:\")\n",
|
|||
|
|
"print(missing_values_after_drop)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"### Можно перейти к созданию выборок"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 444,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Размер обучающей выборки: (2503, 11)\n",
|
|||
|
|
"Размер контрольной выборки: (1074, 11)\n",
|
|||
|
|
"Размер тестовой выборки: (1533, 11)\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Разделение данных на признаки (X) и целевую переменную (y)\n",
|
|||
|
|
"# В данном случае мы хотим предсказать 'stroke'\n",
|
|||
|
|
"X = df.drop(columns=['stroke'])\n",
|
|||
|
|
"y = df['stroke']\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Разбиение данных на обучающую и тестовую выборки\n",
|
|||
|
|
"# Сначала разделим на обучающую и тестовую\n",
|
|||
|
|
"X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Затем разделим обучающую выборку на обучающую и контрольную\n",
|
|||
|
|
"X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.3)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Проверка размеров выборок\n",
|
|||
|
|
"print(\"Размер обучающей выборки:\", X_train.shape)\n",
|
|||
|
|
"print(\"Размер контрольной выборки:\", X_val.shape)\n",
|
|||
|
|
"print(\"Размер тестовой выборки:\", X_test.shape)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"### Оценим сбалансированность выборок:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 445,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Распределение классов в обучающей выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.955653\n",
|
|||
|
|
"1 0.044347\n",
|
|||
|
|
"Name: proportion, dtype: float64\n",
|
|||
|
|
"\n",
|
|||
|
|
"Распределение классов в контрольной выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.954376\n",
|
|||
|
|
"1 0.045624\n",
|
|||
|
|
"Name: proportion, dtype: float64\n",
|
|||
|
|
"\n",
|
|||
|
|
"Распределение классов в тестовой выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.941944\n",
|
|||
|
|
"1 0.058056\n",
|
|||
|
|
"Name: proportion, dtype: float64\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<Figure size 1800x500 with 3 Axes>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Функция для анализа сбалансированности\n",
|
|||
|
|
"def analyze_balance(y_train, y_val, y_test, y_name):\n",
|
|||
|
|
" # Распределение классов\n",
|
|||
|
|
" print(\"Распределение классов в обучающей выборке:\")\n",
|
|||
|
|
" print(y_train.value_counts(normalize=True))\n",
|
|||
|
|
" \n",
|
|||
|
|
" print(\"\\nРаспределение классов в контрольной выборке:\")\n",
|
|||
|
|
" print(y_val.value_counts(normalize=True))\n",
|
|||
|
|
" \n",
|
|||
|
|
" print(\"\\nРаспределение классов в тестовой выборке:\")\n",
|
|||
|
|
" print(y_test.value_counts(normalize=True))\n",
|
|||
|
|
"\n",
|
|||
|
|
" # Создание фигуры и осей для трех столбчатых диаграмм\n",
|
|||
|
|
" fig, axes = plt.subplots(1, 3, figsize=(18, 5), sharey=True)\n",
|
|||
|
|
" fig.suptitle('Распределение в различных выборках')\n",
|
|||
|
|
"\n",
|
|||
|
|
" # Обучающая выборка\n",
|
|||
|
|
" sns.barplot(x=y_train.value_counts().index, y=y_train.value_counts(normalize=True), ax=axes[0])\n",
|
|||
|
|
" axes[0].set_title('Обучающая выборка')\n",
|
|||
|
|
" axes[0].set_xlabel(y_name)\n",
|
|||
|
|
" axes[0].set_ylabel('Доля')\n",
|
|||
|
|
"\n",
|
|||
|
|
" # Контрольная выборка\n",
|
|||
|
|
" sns.barplot(x=y_val.value_counts().index, y=y_val.value_counts(normalize=True), ax=axes[1])\n",
|
|||
|
|
" axes[1].set_title('Контрольная выборка')\n",
|
|||
|
|
" axes[1].set_xlabel(y_name)\n",
|
|||
|
|
"\n",
|
|||
|
|
" # Тестовая выборка\n",
|
|||
|
|
" sns.barplot(x=y_test.value_counts().index, y=y_test.value_counts(normalize=True), ax=axes[2])\n",
|
|||
|
|
" axes[2].set_title('Тестовая выборка')\n",
|
|||
|
|
" axes[2].set_xlabel(y_name)\n",
|
|||
|
|
"\n",
|
|||
|
|
" plt.show()\n",
|
|||
|
|
"\n",
|
|||
|
|
"analyze_balance(y_train, y_val, y_test, 'stroke')"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Легко заметить, что выборки несбалансированны. Необходимо сбалансировать обучающую и контрольную выборки, чтобы получить лучшие результаты при обучении модели. Для балансировки применим RandomOverSampler:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 446,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Распределение классов в обучающей выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.5\n",
|
|||
|
|
"1 0.5\n",
|
|||
|
|
"Name: proportion, dtype: float64\n",
|
|||
|
|
"\n",
|
|||
|
|
"Распределение классов в контрольной выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.5\n",
|
|||
|
|
"1 0.5\n",
|
|||
|
|
"Name: proportion, dtype: float64\n",
|
|||
|
|
"\n",
|
|||
|
|
"Распределение классов в тестовой выборке:\n",
|
|||
|
|
"stroke\n",
|
|||
|
|
"0 0.941944\n",
|
|||
|
|
"1 0.058056\n",
|
|||
|
|
"Name: proportion, dtype: float64\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<Figure size 1800x500 with 3 Axes>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"ros = RandomOverSampler(random_state=42)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Применение RandomOverSampler для балансировки выборок\n",
|
|||
|
|
"X_train_resampled, y_train_resampled = ros.fit_resample(X_train, y_train)\n",
|
|||
|
|
"X_val_resampled, y_val_resampled = ros.fit_resample(X_val, y_val)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Проверка сбалансированности после RandomOverSampler\n",
|
|||
|
|
"analyze_balance(y_train_resampled, y_val_resampled, y_test, 'stroke')"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Выборки сбалансированы.\n",
|
|||
|
|
"\n",
|
|||
|
|
"### Перейдем к конструированию признаков\n",
|
|||
|
|
"\n",
|
|||
|
|
"Для начала применим унитарное кодирование категориальных признаков (one-hot encoding), переведя их в бинарные вектора:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 447,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
" id age hypertension heart_disease avg_glucose_level bmi \\\n",
|
|||
|
|
"0 16605 57.0 0 0 106.24 32.3 \n",
|
|||
|
|
"1 12015 14.0 0 0 99.87 25.2 \n",
|
|||
|
|
"2 26474 44.0 0 0 97.16 33.1 \n",
|
|||
|
|
"3 31143 22.0 0 0 107.52 41.6 \n",
|
|||
|
|
"4 2447 63.0 0 0 85.04 29.7 \n",
|
|||
|
|
"\n",
|
|||
|
|
" gender_Male gender_Other ever_married_Yes work_type_Never_worked \\\n",
|
|||
|
|
"0 True False True False \n",
|
|||
|
|
"1 True False False False \n",
|
|||
|
|
"2 False False True False \n",
|
|||
|
|
"3 False False False False \n",
|
|||
|
|
"4 False False True False \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type_Private work_type_Self-employed work_type_children \\\n",
|
|||
|
|
"0 True False False \n",
|
|||
|
|
"1 False False True \n",
|
|||
|
|
"2 False False False \n",
|
|||
|
|
"3 True False False \n",
|
|||
|
|
"4 True False False \n",
|
|||
|
|
"\n",
|
|||
|
|
" Residence_type_Urban smoking_status_formerly smoked \\\n",
|
|||
|
|
"0 True False \n",
|
|||
|
|
"1 True False \n",
|
|||
|
|
"2 True False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 True True \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_never smoked smoking_status_smokes \n",
|
|||
|
|
"0 True False \n",
|
|||
|
|
"1 False False \n",
|
|||
|
|
"2 False False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 False False \n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Определение категориальных признаков\n",
|
|||
|
|
"categorical_features = ['gender', 'ever_married', 'work_type', 'Residence_type', 'smoking_status']\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Применение one-hot encoding к обучающей выборке\n",
|
|||
|
|
"X_train_encoded = pd.get_dummies(X_train_resampled, columns=categorical_features, drop_first=True)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Применение one-hot encoding к контрольной выборке\n",
|
|||
|
|
"X_val_encoded = pd.get_dummies(X_val_resampled, columns=categorical_features, drop_first=True)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Применение one-hot encoding к тестовой выборке\n",
|
|||
|
|
"X_test_encoded = pd.get_dummies(X_test, columns=categorical_features, drop_first=True)\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(X_train_encoded.head())"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Далее к числовым признакам, а именно к колонке age, применим дискретизацию (позволяет преобразовать данные из числового представления в категориальное):"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 448,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
" id hypertension heart_disease avg_glucose_level bmi gender_Male \\\n",
|
|||
|
|
"0 16605 0 0 106.24 32.3 True \n",
|
|||
|
|
"1 12015 0 0 99.87 25.2 True \n",
|
|||
|
|
"2 26474 0 0 97.16 33.1 False \n",
|
|||
|
|
"3 31143 0 0 107.52 41.6 False \n",
|
|||
|
|
"4 2447 0 0 85.04 29.7 False \n",
|
|||
|
|
"\n",
|
|||
|
|
" gender_Other ever_married_Yes work_type_Never_worked work_type_Private \\\n",
|
|||
|
|
"0 False True False True \n",
|
|||
|
|
"1 False False False False \n",
|
|||
|
|
"2 False True False False \n",
|
|||
|
|
"3 False False False True \n",
|
|||
|
|
"4 False True False True \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type_Self-employed work_type_children Residence_type_Urban \\\n",
|
|||
|
|
"0 False False True \n",
|
|||
|
|
"1 False True True \n",
|
|||
|
|
"2 False False True \n",
|
|||
|
|
"3 False False False \n",
|
|||
|
|
"4 False False True \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_formerly smoked smoking_status_never smoked \\\n",
|
|||
|
|
"0 False True \n",
|
|||
|
|
"1 False False \n",
|
|||
|
|
"2 False False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 True False \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_smokes age_bin \n",
|
|||
|
|
"0 False old \n",
|
|||
|
|
"1 False young \n",
|
|||
|
|
"2 False middle-aged \n",
|
|||
|
|
"3 False young \n",
|
|||
|
|
"4 False old \n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Определение числовых признаков для дискретизации\n",
|
|||
|
|
"numerical_features = ['age']\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Функция для дискретизации числовых признаков\n",
|
|||
|
|
"def discretize_features(df, features, bins, labels):\n",
|
|||
|
|
" for feature in features:\n",
|
|||
|
|
" df[f'{feature}_bin'] = pd.cut(df[feature], bins=bins, labels=labels)\n",
|
|||
|
|
" df.drop(columns=[feature], inplace=True)\n",
|
|||
|
|
" return df\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Заданные интервалы и метки\n",
|
|||
|
|
"age_bins = [0, 25, 55, 100]\n",
|
|||
|
|
"age_labels = [\"young\", \"middle-aged\", \"old\"]\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Применение дискретизации к обучающей, контрольной и тестовой выборкам\n",
|
|||
|
|
"X_train_encoded = discretize_features(X_train_encoded, numerical_features, bins=age_bins, labels=age_labels)\n",
|
|||
|
|
"X_val_encoded = discretize_features(X_val_encoded, numerical_features, bins=age_bins, labels=age_labels)\n",
|
|||
|
|
"X_test_encoded = discretize_features(X_test_encoded, numerical_features, bins=age_bins, labels=age_labels)\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(X_train_encoded.head())"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Применим ручной синтез признаков. Это создание новых признаков на основе существующих, учитывая экспертные знания и логику предметной области. К примеру, в этом случае можно создать признак, в котором вычисляется насколько уровень глюкозы отклоняется от среднего для возрастной группы пациента. Такой признак может быть полезен для выделения пациентов с нетипичными данными."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 449,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
" id hypertension heart_disease avg_glucose_level bmi gender_Male \\\n",
|
|||
|
|
"0 16605 0 0 106.24 32.3 True \n",
|
|||
|
|
"1 12015 0 0 99.87 25.2 True \n",
|
|||
|
|
"2 26474 0 0 97.16 33.1 False \n",
|
|||
|
|
"3 31143 0 0 107.52 41.6 False \n",
|
|||
|
|
"4 2447 0 0 85.04 29.7 False \n",
|
|||
|
|
"\n",
|
|||
|
|
" gender_Other ever_married_Yes work_type_Never_worked work_type_Private \\\n",
|
|||
|
|
"0 False True False True \n",
|
|||
|
|
"1 False False False False \n",
|
|||
|
|
"2 False True False False \n",
|
|||
|
|
"3 False False False True \n",
|
|||
|
|
"4 False True False True \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type_Self-employed work_type_children Residence_type_Urban \\\n",
|
|||
|
|
"0 False False True \n",
|
|||
|
|
"1 False True True \n",
|
|||
|
|
"2 False False True \n",
|
|||
|
|
"3 False False False \n",
|
|||
|
|
"4 False False True \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_formerly smoked smoking_status_never smoked \\\n",
|
|||
|
|
"0 False True \n",
|
|||
|
|
"1 False False \n",
|
|||
|
|
"2 False False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 True False \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_smokes age_bin glucose_age_deviation \n",
|
|||
|
|
"0 False old -27.642870 \n",
|
|||
|
|
"1 False young 6.088032 \n",
|
|||
|
|
"2 False middle-aged -6.217053 \n",
|
|||
|
|
"3 False young 13.738032 \n",
|
|||
|
|
"4 False old -48.842870 \n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"age_glucose_mean = X_train_encoded.groupby('age_bin', observed=False)['avg_glucose_level'].transform('mean')\n",
|
|||
|
|
"X_train_encoded['glucose_age_deviation'] = X_train_encoded['avg_glucose_level'] - age_glucose_mean\n",
|
|||
|
|
"\n",
|
|||
|
|
"age_glucose_mean = X_val_encoded.groupby('age_bin', observed=False)['avg_glucose_level'].transform('mean')\n",
|
|||
|
|
"X_val_encoded['glucose_age_deviation'] = X_val_encoded['avg_glucose_level'] - age_glucose_mean\n",
|
|||
|
|
"\n",
|
|||
|
|
"age_glucose_mean = X_test_encoded.groupby('age_bin', observed=False)['avg_glucose_level'].transform('mean')\n",
|
|||
|
|
"X_test_encoded['glucose_age_deviation'] = X_test_encoded['avg_glucose_level'] - age_glucose_mean\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(X_train_encoded.head())"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Теперь используем масштабирование признаков, что позволяет привести все числовые признаки к одинаковым или очень похожим диапазонам значений либо распределениям. По результатам многочисленных исследований масштабирование признаков позволяет получить более качественную модель за счет снижения доминирования одних признаков над другими."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 450,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
" id hypertension heart_disease avg_glucose_level bmi \\\n",
|
|||
|
|
"0 16605 0 0 -0.244097 0.426328 \n",
|
|||
|
|
"1 12015 0 0 -0.360110 -0.596170 \n",
|
|||
|
|
"2 26474 0 0 -0.409465 0.541539 \n",
|
|||
|
|
"3 31143 0 0 -0.220785 1.765656 \n",
|
|||
|
|
"4 2447 0 0 -0.630199 0.051892 \n",
|
|||
|
|
"\n",
|
|||
|
|
" gender_Male gender_Other ever_married_Yes work_type_Never_worked \\\n",
|
|||
|
|
"0 True False True False \n",
|
|||
|
|
"1 True False False False \n",
|
|||
|
|
"2 False False True False \n",
|
|||
|
|
"3 False False False False \n",
|
|||
|
|
"4 False False True False \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type_Private work_type_Self-employed work_type_children \\\n",
|
|||
|
|
"0 True False False \n",
|
|||
|
|
"1 False False True \n",
|
|||
|
|
"2 False False False \n",
|
|||
|
|
"3 True False False \n",
|
|||
|
|
"4 True False False \n",
|
|||
|
|
"\n",
|
|||
|
|
" Residence_type_Urban smoking_status_formerly smoked \\\n",
|
|||
|
|
"0 True False \n",
|
|||
|
|
"1 True False \n",
|
|||
|
|
"2 True False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 True True \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_never smoked smoking_status_smokes age_bin \\\n",
|
|||
|
|
"0 True False old \n",
|
|||
|
|
"1 False False young \n",
|
|||
|
|
"2 False False middle-aged \n",
|
|||
|
|
"3 False False young \n",
|
|||
|
|
"4 False False old \n",
|
|||
|
|
"\n",
|
|||
|
|
" glucose_age_deviation \n",
|
|||
|
|
"0 -0.528807 \n",
|
|||
|
|
"1 0.116464 \n",
|
|||
|
|
"2 -0.118932 \n",
|
|||
|
|
"3 0.262808 \n",
|
|||
|
|
"4 -0.934362 \n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Пример масштабирования числовых признаков\n",
|
|||
|
|
"numerical_features = ['avg_glucose_level', 'bmi', 'glucose_age_deviation']\n",
|
|||
|
|
"\n",
|
|||
|
|
"scaler = StandardScaler()\n",
|
|||
|
|
"X_train_encoded[numerical_features] = scaler.fit_transform(X_train_encoded[numerical_features])\n",
|
|||
|
|
"X_val_encoded[numerical_features] = scaler.transform(X_val_encoded[numerical_features])\n",
|
|||
|
|
"X_test_encoded[numerical_features] = scaler.transform(X_test_encoded[numerical_features])\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(X_train_encoded.head())"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"И также попробуем сконструировать признаки, используя фреймворк Featuretools:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 451,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
" id hypertension heart_disease avg_glucose_level bmi \\\n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 16605 0 0 -0.244097 0.426328 \n",
|
|||
|
|
"1 12015 0 0 -0.360110 -0.596170 \n",
|
|||
|
|
"2 26474 0 0 -0.409465 0.541539 \n",
|
|||
|
|
"3 31143 0 0 -0.220785 1.765656 \n",
|
|||
|
|
"4 2447 0 0 -0.630199 0.051892 \n",
|
|||
|
|
"\n",
|
|||
|
|
" gender_Male gender_Other ever_married_Yes work_type_Never_worked \\\n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 True False True False \n",
|
|||
|
|
"1 True False False False \n",
|
|||
|
|
"2 False False True False \n",
|
|||
|
|
"3 False False False False \n",
|
|||
|
|
"4 False False True False \n",
|
|||
|
|
"\n",
|
|||
|
|
" work_type_Private work_type_Self-employed work_type_children \\\n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 True False False \n",
|
|||
|
|
"1 False False True \n",
|
|||
|
|
"2 False False False \n",
|
|||
|
|
"3 True False False \n",
|
|||
|
|
"4 True False False \n",
|
|||
|
|
"\n",
|
|||
|
|
" Residence_type_Urban smoking_status_formerly smoked \\\n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 True False \n",
|
|||
|
|
"1 True False \n",
|
|||
|
|
"2 True False \n",
|
|||
|
|
"3 False False \n",
|
|||
|
|
"4 True True \n",
|
|||
|
|
"\n",
|
|||
|
|
" smoking_status_never smoked smoking_status_smokes age_bin \\\n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 True False old \n",
|
|||
|
|
"1 False False young \n",
|
|||
|
|
"2 False False middle-aged \n",
|
|||
|
|
"3 False False young \n",
|
|||
|
|
"4 False False old \n",
|
|||
|
|
"\n",
|
|||
|
|
" glucose_age_deviation \n",
|
|||
|
|
"index \n",
|
|||
|
|
"0 -0.528807 \n",
|
|||
|
|
"1 0.116464 \n",
|
|||
|
|
"2 -0.118932 \n",
|
|||
|
|
"3 0.262808 \n",
|
|||
|
|
"4 -0.934362 \n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"name": "stderr",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"c:\\Users\\Ilya\\Desktop\\AIM\\aimenv\\Lib\\site-packages\\woodwork\\type_sys\\utils.py:33: UserWarning: Could not infer format, so each element will be parsed individually, falling back to `dateutil`. To ensure parsing is consistent and as-expected, please specify a format.\n",
|
|||
|
|
" pd.to_datetime(\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"data = X_train_encoded.copy() # Используем предобработанные данные\n",
|
|||
|
|
"\n",
|
|||
|
|
"es = ft.EntitySet(id=\"patients\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"es = es.add_dataframe(dataframe_name=\"strokes_data\", dataframe=data, index=\"index\", make_index=True)\n",
|
|||
|
|
"\n",
|
|||
|
|
"feature_matrix, feature_defs = ft.dfs(\n",
|
|||
|
|
" entityset=es, \n",
|
|||
|
|
" target_dataframe_name=\"strokes_data\",\n",
|
|||
|
|
" max_depth=1\n",
|
|||
|
|
")\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(feature_matrix.head())"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"### Оценим качество набора признаков.\n",
|
|||
|
|
"\n",
|
|||
|
|
"Представим основные оценки качества наборов признаков: \n",
|
|||
|
|
"\n",
|
|||
|
|
"* Предсказательная способность Метрики: RMSE, MAE, R²\n",
|
|||
|
|
"\n",
|
|||
|
|
" Методы: Обучение модели на обучающей выборке и оценка на контрольной и тестовой выборках.\n",
|
|||
|
|
"\n",
|
|||
|
|
"* Скорость вычисления \n",
|
|||
|
|
"\n",
|
|||
|
|
" Методы: Измерение времени выполнения генерации признаков и обучения модели.\n",
|
|||
|
|
"\n",
|
|||
|
|
"* Надежность \n",
|
|||
|
|
"\n",
|
|||
|
|
" Методы: Кросс-валидация, анализ чувствительности модели к изменениям в данных.\n",
|
|||
|
|
"\n",
|
|||
|
|
"* Корреляция \n",
|
|||
|
|
"\n",
|
|||
|
|
" Методы: Анализ корреляционной матрицы признаков, удаление мультиколлинеарных признаков.\n",
|
|||
|
|
"\n",
|
|||
|
|
"* Цельность \n",
|
|||
|
|
"\n",
|
|||
|
|
" Методы: Проверка логической связи между признаками и целевой переменной, интерпретация результатов модели."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 452,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Время обучения модели: 0.01 секунд\n",
|
|||
|
|
"Среднеквадратичная ошибка: 0.41\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"X_train_encoded = pd.get_dummies(X_train_encoded, drop_first=True)\n",
|
|||
|
|
"X_val_encoded = pd.get_dummies(X_val_encoded, drop_first=True)\n",
|
|||
|
|
"X_test_encoded = pd.get_dummies(X_test_encoded, drop_first=True)\n",
|
|||
|
|
"\n",
|
|||
|
|
"all_columns = X_train_encoded.columns\n",
|
|||
|
|
"X_train_encoded = X_train_encoded.reindex(columns=all_columns, fill_value=0)\n",
|
|||
|
|
"X_val_encoded = X_val_encoded.reindex(columns=all_columns, fill_value=0)\n",
|
|||
|
|
"X_test_encoded = X_test_encoded.reindex(columns=all_columns, fill_value=0)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Обучение модели\n",
|
|||
|
|
"model = LinearRegression()\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Начинаем отсчет времени\n",
|
|||
|
|
"start_time = time.time()\n",
|
|||
|
|
"model.fit(X_train_encoded, y_train_resampled)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Время обучения модели\n",
|
|||
|
|
"train_time = time.time() - start_time\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Предсказания и оценка модели и вычисляем среднеквадратичную ошибку\n",
|
|||
|
|
"predictions = model.predict(X_val_encoded)\n",
|
|||
|
|
"mse = root_mean_squared_error(y_val_resampled, predictions)\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(f'Время обучения модели: {train_time:.2f} секунд')\n",
|
|||
|
|
"print(f'Среднеквадратичная ошибка: {mse:.2f}')"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 453,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"\n",
|
|||
|
|
"RMSE: 0.24109840514907446\n",
|
|||
|
|
"R²: -0.06295721700021817\n",
|
|||
|
|
"MAE: 0.10402478799739073 \n",
|
|||
|
|
"\n",
|
|||
|
|
"Кросс-валидация RMSE: 0.1197518340742331 \n",
|
|||
|
|
"\n",
|
|||
|
|
"Train RMSE: 0.037396456827854585\n",
|
|||
|
|
"Train R²: 0.9944060200668896\n",
|
|||
|
|
"Train MAE: 0.010727424749163881\n",
|
|||
|
|
"\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"# Выбор модели\n",
|
|||
|
|
"model = RandomForestRegressor(random_state=42)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Обучение модели\n",
|
|||
|
|
"model.fit(X_train_encoded, y_train_resampled)\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Предсказание и оценка\n",
|
|||
|
|
"y_pred = model.predict(X_test_encoded)\n",
|
|||
|
|
"\n",
|
|||
|
|
"rmse = root_mean_squared_error(y_test, y_pred)\n",
|
|||
|
|
"r2 = r2_score(y_test, y_pred)\n",
|
|||
|
|
"mae = mean_absolute_error(y_test, y_pred)\n",
|
|||
|
|
"\n",
|
|||
|
|
"print()\n",
|
|||
|
|
"print(f\"RMSE: {rmse}\")\n",
|
|||
|
|
"print(f\"R²: {r2}\")\n",
|
|||
|
|
"print(f\"MAE: {mae} \\n\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Кросс-валидация\n",
|
|||
|
|
"scores = cross_val_score(model, X_train_encoded, y_train_resampled, cv=5, scoring='neg_mean_squared_error')\n",
|
|||
|
|
"rmse_cv = (-scores.mean())**0.5\n",
|
|||
|
|
"print(f\"Кросс-валидация RMSE: {rmse_cv} \\n\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"# Проверка на переобучение\n",
|
|||
|
|
"y_train_pred = model.predict(X_train_encoded)\n",
|
|||
|
|
"\n",
|
|||
|
|
"rmse_train = root_mean_squared_error(y_train_resampled, y_train_pred)\n",
|
|||
|
|
"r2_train = r2_score(y_train_resampled, y_train_pred)\n",
|
|||
|
|
"mae_train = mean_absolute_error(y_train_resampled, y_train_pred)\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(f\"Train RMSE: {rmse_train}\")\n",
|
|||
|
|
"print(f\"Train R²: {r2_train}\")\n",
|
|||
|
|
"print(f\"Train MAE: {mae_train}\")\n",
|
|||
|
|
"print()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {},
|
|||
|
|
"source": [
|
|||
|
|
"Можно заметить, что модель хорошо подстроилась под тренировочные данные (Низкий Train RMSE и высокое значение Train R²). Однако высокий RMSE и отрицательный R² на тестовом наборе свидетельствуют о том, что модель не обобщила зависимости и плохо предсказывает новые данные, поэтому можно сделать вывод о том, что получившийся набор признаков, к сожалению, далек от идеала. "
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"metadata": {
|
|||
|
|
"kernelspec": {
|
|||
|
|
"display_name": "aimenv",
|
|||
|
|
"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": "3.12.5"
|
|||
|
|
}
|
|||
|
|
},
|
|||
|
|
"nbformat": 4,
|
|||
|
|
"nbformat_minor": 2
|
|||
|
|
}
|