KFold交叉验证实现

实现代码 

import pandas as pd
from sklearn.model_selection import GroupKFold

train = pd.read_csv("../../Data/train.csv")  # (6079, 41)
test = pd.read_csv("../../Data/test.csv")  # (476, 11)


num_folds = 5
gkf = GroupKFold(n_splits=num_folds).split(X=train.question_body, groups=train.question_body)  # train.question_body 用来获取总长度
print(gkf)


for fold, (train_idx, valid_idx) in enumerate(gkf):  # 五折交叉验证
    print("fold:{}, 训练集长度:{}, 验证集长度：{}".format(fold, len(train_idx), len(valid_idx)))

    train_data = train.iloc[train_idx]  # 按照行索引 获取数据
    valid_data = train.iloc[valid_idx]  # 验证集

    print(len(train_data))
    print(len(valid_data))
    assert len(train_data) + len(valid_data) == len(train), ("数据划分不正确!!!",)