sklearn 抽样

1 交叉验证包中的train_test_split

设置测试集比例，在原始数据中随机采样，但是所得样本中各个类别比例保持与原样本一致。如下例所示

import numpy as npdata_x=[['this is class 1 ']]*100 + [['this is class 2']]*50data_y=[[1]]*100 + [[2]]*50data_yX_train, X_test, y_train, y_test = train_test_split(data_x,data_y,test_size = 0.3)X_train#sum(y_train==1)/len(y_train)print('numbers of positive class in training data:', sum( np.mat(y_train)==1 )[0],'/',len(y_train) )print('numbers of negative class in training data:', sum( np.mat(y_train)==2 )[0],'/', len(y_train))print('numbers of positive class in test data:',sum(np.mat(y_test)== 1 ),'/',len(y_test))

numbers of positive class in training set: [[74]] / 105
numbers of negative class in training set: [[31]] / 105
numbers of positive class in test data: [[26]] / 45

待更新