106 lines
4.3 KiB
Python
106 lines
4.3 KiB
Python
|
import numpy as np
|
|||
|
|
import pandas as pd
|
|||
|
|
from sklearn.model_selection import train_test_split
|
|||
|
|
from sklearn.linear_model import LinearRegression
|
|||
|
|
from sklearn.ensemble import RandomForestRegressor
|
|||
|
|
from sklearn.feature_selection import f_regression
|
|||
|
|
from sklearn.preprocessing import MinMaxScaler
|
|||
|
|
|
|||
|
|
# загрузка dataset
|
|||
|
|
data = pd.read_csv('dataset.csv')
|
|||
|
|
|
|||
|
|
# разделение dataset на тренировочную и тестовую выборки
|
|||
|
|
X = data.drop(['Target'], axis=1)
|
|||
|
|
y = data['Target']
|
|||
|
|
|
|||
|
|
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
|
|||
|
|
|
|||
|
|
# Тренировка моделей
|
|||
|
|
# Линейная регрессия
|
|||
|
|
lr = LinearRegression()
|
|||
|
|
lr.fit(X_train, y_train)
|
|||
|
|
|
|||
|
|
# Сокращение признаков случайными деревьями с помощью Random Forest Regressor
|
|||
|
|
rf = RandomForestRegressor()
|
|||
|
|
rf.fit(X_train, y_train)
|
|||
|
|
|
|||
|
|
# Ранжирование признаков использую каждую модель/метод
|
|||
|
|
# Получение абсолютных значений коэффициентов в качестве оценок важности признаков
|
|||
|
|
lr_scores = abs(lr.coef_)
|
|||
|
|
|
|||
|
|
# Получение оценок важности объектов из модели Random Forest Regressor
|
|||
|
|
rf_scores = rf.feature_importances_
|
|||
|
|
|
|||
|
|
# Отображение итоговых оценок по каждой колонке
|
|||
|
|
feature_names = X.columns.tolist()
|
|||
|
|
|
|||
|
|
# показать оценки рангов по модели линейной регрессии
|
|||
|
|
print("оценки линейной регрессии:")
|
|||
|
|
for feature, score in zip(feature_names, lr_scores):
|
|||
|
|
print(f"{feature}: {round(score, 4)}")
|
|||
|
|
|
|||
|
|
# оценки метода рандомных лесов
|
|||
|
|
print("\nоценки Random Forest:")
|
|||
|
|
for feature, score in zip(feature_names, rf_scores):
|
|||
|
|
print(f"{feature}: {round(score, 4)}")
|
|||
|
|
|
|||
|
|
# вычисление значений оценки для f_regression
|
|||
|
|
f_scores, p_values = f_regression(X, y)
|
|||
|
|
|
|||
|
|
# оценки f_regression
|
|||
|
|
print("\nоценки f_regression:")
|
|||
|
|
for feature, score in zip(feature_names, f_scores):
|
|||
|
|
print(f"{feature}: {round(score, 4)}")
|
|||
|
|
|
|||
|
|
# использую MinMaxScaler для точных средних значений рангов
|
|||
|
|
scaler = MinMaxScaler()
|
|||
|
|
lr_scores_scaled = scaler.fit_transform(lr_scores.reshape(-1, 1)).flatten()
|
|||
|
|
rf_scores_scaled = scaler.fit_transform(rf_scores.reshape(-1, 1)).flatten()
|
|||
|
|
f_scores_scaled = scaler.fit_transform(f_scores.reshape(-1, 1)).flatten()
|
|||
|
|
|
|||
|
|
# вычисление средних оценок для каждого признака
|
|||
|
|
average_scores = {}
|
|||
|
|
for feature in feature_names:
|
|||
|
|
average_scores[feature] = (lr_scores_scaled[feature_names.index(feature)] +
|
|||
|
|
rf_scores_scaled[feature_names.index(feature)] +
|
|||
|
|
f_scores_scaled[feature_names.index(feature)]) / 3
|
|||
|
|
|
|||
|
|
# получаем среднюю оценку признаков
|
|||
|
|
sorted_features = sorted(average_scores.items(), key=lambda x: x[1], reverse=True)
|
|||
|
|
|
|||
|
|
# получаем самых важных признака
|
|||
|
|
top_4_features = sorted_features[:4]
|
|||
|
|
|
|||
|
|
# отображаем 4 самые важные
|
|||
|
|
print("\n4 самых важных признака в среднем:")
|
|||
|
|
for feature, score in top_4_features:
|
|||
|
|
print(f"Признак: {feature}, Оценка: {round(score, 4)}")
|
|||
|
|
|
|||
|
|
|
|||
|
|
# отображаем самых важных признака для каждого метода/модели
|
|||
|
|
top_lr_indices = np.argsort(lr_scores)[-4:][::-1]
|
|||
|
|
top_rf_indices = np.argsort(rf_scores)[-4:][::-1]
|
|||
|
|
top_f_indices = np.argsort(f_scores)[-4:][::-1]
|
|||
|
|
|
|||
|
|
top_lr_features = [feature_names[i] for i in top_lr_indices]
|
|||
|
|
top_rf_features = [feature_names[i] for i in top_rf_indices]
|
|||
|
|
top_f_features = [feature_names[i] for i in top_f_indices]
|
|||
|
|
|
|||
|
|
top_lr_features_score = [lr_scores[i] for i in top_lr_indices]
|
|||
|
|
top_rf_features_score = [rf_scores[i] for i in top_rf_indices]
|
|||
|
|
top_f_features_score = [f_scores[i] for i in top_f_indices]
|
|||
|
|
|
|||
|
|
print("\n4 самых важных для lr_scores:")
|
|||
|
|
print(top_lr_features)
|
|||
|
|
for i in top_lr_features_score:
|
|||
|
|
print(round(i, 4))
|
|||
|
|
|
|||
|
|
print("\n4 самых важных для rf_scores:")
|
|||
|
|
print(top_rf_features)
|
|||
|
|
for i in top_rf_features_score:
|
|||
|
|
print(round(i, 4))
|
|||
|
|
|
|||
|
|
print("\n4 самых важных для f_scores:")
|
|||
|
|
print(top_f_features)
|
|||
|
|
for i in top_f_features_score:
|
|||
|
|
print(round(i, 4))
|