Python中进行特征重要性分析的9个常用方法_利用python对模型的特征重要程度进行分析-优快云博客

特征重要性分析用于了解每个特征(变量或输入)对于做出预测的有用性或价值。目标是确定对模型输出影响最大的最重要的特征，它是机器学习中经常使用的一种方法。

为什么特征重要性分析很重要?

如果有一个包含数十个甚至数百个特征的数据集，每个特征都可能对你的机器学习模型的性能有所贡献。但是并不是所有的特征都是一样的。有些可能是冗余的或不相关的，这会增加建模的复杂性并可能导致过拟合。

特征重要性分析可以识别并关注最具信息量的特征，从而带来以下几个优势:

改进的模型性能
减少过度拟合
更快的训练和推理
增强的可解释性

下面我们深入了解在Python中的一些特性重要性分析的方法。

特征重要性分析方法

1、排列重要性 PermutationImportance

该方法会随机排列每个特征的值，然后监控模型性能下降的程度。如果获得了更大的下降意味着特征更重要



from sklearn.datasets import load\_breast\_cancer  
from sklearn.ensemble import RandomForestClassifier  
from sklearn.inspection import permutation\_importance  
from sklearn.model\_selection import train\_test\_split  
import matplotlib.pyplot as plt  
   
cancer = load\_breast\_cancer()  
   
X\_train, X\_test, y\_train, y\_test = train\_test\_split(cancer.data, cancer.target, random\_state=1)  
   
rf = RandomForestClassifier(n\_estimators=100, random\_state=1)  
rf.fit(X\_train, y\_train)  
   
baseline = rf.score(X\_test, y\_test)  
result = permutation\_importance(rf, X\_test, y\_test, n\_repeats=10, random\_state=1, scoring='accuracy')  
   
importances = result.importances\_mean  
   
# Visualize permutation importances  
plt.bar(range(len(importances)), importances)  
plt.xlabel('Feature Index')  
plt.ylabel('Permutation Importance')  
plt.show()

[外链图片转存中…(img-pi8NSf8x-1704351889783)]

2、内置特征重要性(coef_或feature_importances_)

一些模型，如线性回归和随机森林，可以直接输出特征重要性分数。这些显示了每个特征对最终预测的贡献。



from sklearn.datasets import load\_breast\_cancer  
from sklearn.ensemble import RandomForestClassifier  
   
X, y = load\_breast\_cancer(return\_X\_y=True)  
   
rf = RandomForestClassifier(n\_estimators=100, random\_state=1)  
rf.fit(X, y)  
   
importances = rf.feature\_importances\_  
   
# Plot importances  
plt.bar(range(X.shape\[1\]), importances)  
plt.xlabel('Feature Index')  
plt.ylabel('Feature Importance')  
plt.show()

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

3、Leave-one-out

迭代地每次删除一个特征并评估准确性。



 from sklearn.datasets import load\_breast\_cancer  
 from sklearn.model\_selection import train\_test\_split  
 from sklearn.ensemble import RandomForestClassifier  
 from sklearn.metrics import accuracy\_score  
 import matplotlib.pyplot as plt  
 import numpy as np  
   
 # Load sample data  
 X, y = load\_breast\_cancer(return\_X\_y=True)  
   
 # Split data into train and test sets  
 X\_train, X\_test, y\_train, y\_test = train\_test\_split(X, y, test\_size=0.3, random\_state=1)  
   
 # Train a random forest model  
 rf = RandomForestClassifier(n\_estimators=100, random\_state=1)  
 rf.fit(X\_train, y\_train)  
   
 # Get baseline accuracy on test data  
 base\_acc = accuracy\_score(y\_test, rf.predict(X\_test))  
   
 # Initialize empty list to store importances  
 importances = \[\]  
   
 # Iterate over all columns and remove one at a time  
 for i in range(X\_train.shape\[1\]):  
    X\_temp = np.delete(X\_train, i, axis=1)  
    rf.fit(X\_temp, y\_train)  
    acc = accuracy\_score(y\_test, rf.predict(np.delete(X\_test, i, axis=1)))  
    importances.append(base\_acc - acc)  
       
 # Plot importance scores      
 plt.bar(range(len(importances)), importances)  
 plt.show()

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

4、相关性分析

计算各特征与目标变量之间的相关性。相关性越高的特征越重要。



import pandas as pd  
from sklearn.datasets import load\_breast\_cancer  
   
X, y = load\_breast\_cancer(return\_X\_y=True)  
df = pd.DataFrame(X, columns=range(30))  
df\['y'\] = y  
   
correlations = df.corrwith(df.y).abs()  
correlations.sort\_values(ascending=False, inplace=True)  
   
correlations.plot.bar()

5、递归特征消除 Recursive Feature Elimination

递归地删除特征并查看它如何影响模型性能。删除时会导致更大下降的特征更重要。



 from sklearn.ensemble import RandomForestClassifier  
 from sklearn.feature\_selection import RFE  
 import pandas as pd  
 from sklearn.datasets import load\_breast\_cancer  
 import matplotlib.pyplot as plt  
   
 X, y = load\_breast\_cancer(return\_X\_y=True)  
 df = pd.DataFrame(X, columns=range(30))  
 df\['y'\] = y  
   
 rf = RandomForestClassifier()  
   
 rfe = RFE(rf, n\_features\_to\_select=10)  
 rfe.fit(X, y)  
   
 print(rfe.ranking\_)

输出为[6 4 11 12 7 11 18 21 8 16 10 3 15 14 19 17 20 13 11 11 12 9 11 5 11]

6、XGBoost特性重要性

计算一个特性用于跨所有树拆分数据的次数。更多的分裂意味着更重要。



 import xgboost as xgb  
 import pandas as pd  
 from sklearn.datasets import load\_breast\_cancer  
 import matplotlib.pyplot as plt  
   
 X, y = load\_breast\_cancer(return\_X\_y=True)  
 df = pd.DataFrame(X, columns=range(30))  
 df\['y'\] = y  
   
 model = xgb.XGBClassifier()  
 model.fit(X, y)  
   
 importances = model.feature\_importances\_  
 importances = pd.Series(importances, index=range(X.shape\[1\]))  
 importances.plot.bar()

7、主成分分析 PCA

对特征进行主成分分析，并查看每个主成分的解释方差比。在前几个组件上具有较高负载的特性更为重要。



 from sklearn.decomposition import PCA  
 import pandas as pd  
 from sklearn.datasets import load\_breast\_cancer  
 import matplotlib.pyplot as plt  
   
 X, y = load\_breast\_cancer(return\_X\_y=True)  
 df = pd.DataFrame(X, columns=range(30))  
 df\['y'\] = y  
   
 pca = PCA()  
 pca.fit(X)  
   
 plt.bar(range(pca.n\_components\_), pca.explained\_variance\_ratio\_)  
 plt.xlabel('PCA components')  
 plt.ylabel('Explained Variance')

8、方差分析 ANOVA

使用f_classif()获得每个特征的方差分析f值。f值越高，表明特征与目标的相关性越强。



 from sklearn.feature\_selection import f\_classif  
 import pandas as pd  
 from sklearn.datasets import load\_breast\_cancer  
 import matplotlib.pyplot as plt  
   
 X, y = load\_breast\_cancer(return\_X\_y=True)  
 df = pd.DataFrame(X, columns=range(30))  
 df\['y'\] = y  
   
 fval = f\_classif(X, y)  
 fval = pd.Series(fval\[0\], index=range(X.shape\[1\]))  
 fval.plot.bar()

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传

9、卡方检验

使用chi2()获得每个特征的卡方统计信息。得分越高的特征越有可能独立于目标。



 from sklearn.feature\_selection import chi2  
 import pandas as pd  
 from sklearn.datasets import load\_breast\_cancer  
 import matplotlib.pyplot as plt  
   
 X, y = load\_breast\_cancer(return\_X\_y=True)  
 df = pd.DataFrame(X, columns=range(30))  
 df\['y'\] = y  
   
 chi\_scores = chi2(X, y)  
 chi\_scores = pd.Series(chi\_scores\[0\], index=range(X.shape\[1\]))  
 chi\_scores.plot.bar()

外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传