Python机器学习特征重要性分析的8个常用方法实例探究

from sklearn.tree import DecisionTreeClassifier
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt
# 加载数据集
data = load_iris()
X = data.data
y = data.target
# 构建决策树模型
model = DecisionTreeClassifier()
model.fit(X, y)
# 获取特征重要性
importance = model.feature_importances_
# 特征重要性可视化
plt.barh(range(X.shape[1]), importance, align='center')
plt.yticks(range(X.shape[1]), data.feature_names)
plt.xlabel('Feature Importance')
plt.ylabel('Features')
plt.show()

from sklearn.ensemble import RandomForestClassifier
# 构建Random Forest模型
rf_model = RandomForestClassifier()
rf_model.fit(X, y)
# 获取特征重要性
importance_rf = rf_model.feature_importances_
# 可视化Random Forest的特征重要性
plt.barh(range(X.shape[1]), importance_rf, align='center')
plt.yticks(range(X.shape[1]), data.feature_names)
plt.xlabel('Feature Importance')
plt.ylabel('Features')
plt.show()

import pandas as pd
# 创建DataFrame
df = pd.DataFrame(data.data, columns=data.feature_names)
df['target'] = data.target
# 计算Pearson相关系数
correlation = df.corr()
# 可视化相关系数矩阵
import seaborn as sns
plt.figure(figsize=(10, 8))
sns.heatmap(correlation, annot=True, cmap='coolwarm')
plt.title('Pearson Correlation Matrix')
plt.show()

from sklearn.feature_selection import mutual_info_classif

# 计算互信息
mi = mutual_info_classif(X, y)

# 可视化互信息
plt.barh(range(X.shape[1]), mi, align='center')
plt.yticks(range(X.shape[1]), data.feature_names)
plt.xlabel('Mutual Information')
plt.ylabel('Features')
plt.show()

import shap

# 创建并训练一个模型（例如XGBoost）
model = xgb.XGBClassifier()
model.fit(X, y)

# 创建一个SHAP解释器
explainer = shap.Explainer(model)
shap_values = explainer.shap_values(X)

# 可视化SHAP值
shap.summary_plot(shap_values, X, feature_names=data.feature_names, plot_type="bar")

from sklearn.inspection import permutation_importance

# 计算Permutation Feature Importance
result = permutation_importance(model, X, y, n_repeats=10, random_state=42)

# 可视化Permutation Feature Importance
sorted_idx = result.importances_mean.argsort()
plt.barh(range(X.shape[1]), result.importances_mean[sorted_idx], align='center')
plt.yticks(range(X.shape[1]), data.feature_names[sorted_idx])
plt.xlabel('Permutation Importance')
plt.ylabel('Features')
plt.show()

from sklearn.ensemble import GradientBoostingClassifier

# 构建GBDT模型
gbdt_model = GradientBoostingClassifier()
gbdt_model.fit(X, y)

# 获取特征重要性
importance_gbdt = gbdt_model.feature_importances_

# 可视化GBDT的特征重要性
plt.barh(range(X.shape[1]), importance_gbdt, align='center')
plt.yticks(range(X.shape[1]), data.feature_names)
plt.xlabel('Feature Importance')
plt.ylabel('Features')
plt.show()

import xgboost as xgb
# 转换数据为DMatrix格式
dtrain = xgb.DMatrix(X, label=y)
# 定义参数
param = {'objective': 'multi:softmax', 'num_class': 3}
# 训练模型
num_round = 10
xgb_model = xgb.train(param, dtrain, num_round)
# 可视化特征重要性
xgb.plot_importance(xgb_model)
plt.show()