imblearn 管道框架中的数据预处理步骤答案

【问题标题】：Data pre-processing steps within the imblearn pipeline frameworkimblearn 管道框架中的数据预处理步骤
【发布时间】：2021-07-15 01:24:33
【问题描述】：

如何在以下imblearn.pipeline 管道中包含SimpleImputer(strategy='constant',fill_value= 0)) 和("scale", MaxAbsScaler())？我看不到在imblearn.pipeline 管道框架中包含数据预处理步骤的任何选项。非常感谢您的建议！

smote_pipe = make_imb_pipeline(SMOTE(), LogisticRegression())
scores = cross_validate(smote_pipe, X_train, y_train, cv=10,
                        scoring=('roc_auc', 'average_precision'))
pd.DataFrame(scores)[['test_roc_auc', 'test_average_precision']].mean()

【问题讨论】：

标签： machine-learning scikit-learn classification imbalanced-data imblearn

【解决方案1】：

文档中可能没有提到它们，但您可以包含它们。

这是一个相当简单的可重现示例：

import numpy as np
from imblearn.over_sampling import SMOTE
from imblearn.pipeline import make_pipeline as make_imb_pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import MinMaxScaler
from sklearn.model_selection import cross_validate
from sklearn.linear_model import LogisticRegression


X = np.array([[-0.718,np.nan,-0.626,np.nan,-0.286,-0.262,-0.526,-0.662,-0.578,-0.418,-0.374,-0.482,-0.678,-0.678,-0.562,-0.362,-0.346,-0.562,-0.442,-0.466,-0.434,-0.314,-0.110,-0.146,-0.390,-0.614,-0.262,-0.310,-0.458,-0.350,-0.190,-0.0940,-0.0860,-0.182,-0.190,-0.170,-0.298,-0.562,-0.742,-0.750,-0.566,-0.438,0.0660,-0.0300,-0.0660,0.0300,-0.0220,-0.142,-0.122,0.0460,-0.0980,0.0300,0.138,0.150],[0.279,0.439,0.643,0.715,0.555,0.347,0.295,0.431,0.491,0.507,0.407,0.403,0.483,0.599,0.771,0.747,0.559,0.627,0.763,0.519,0.695,0.723,0.563,0.227,0.175,0.211,0.259,0.275,0.223,0.103,0.143,0.347,0.563,0.715,0.399,0.615,0.811,0.791,0.611,0.315,0.239,0.123,0.395,0.399,0.531,0.515,0.471,0.523,0.663,0.651,0.427,0.347,0.443,0.587]]).reshape(-1, 2)
y = np.array([0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1,1,1])


oversample_pipe = make_imb_pipeline(
    SimpleImputer(strategy='constant', fill_value=0),
    MinMaxScaler(),
    SMOTE(),
    LogisticRegression()
)
scores = cross_validate(
    oversample_pipe, X, y, cv=5, scoring=("roc_auc", "average_precision")
)
print(scores["test_roc_auc"].mean())
print(scores["test_average_precision"].mean())

输出（粗略地说，有一些随机性）：

0.8375
0.7485714285714286

【讨论】：