python 分層抽樣

import numpy as np
import pandas as pd

PATH_DES = '/Users/linxianli/Desktop/'
df = pd.read_excel(PATH_DES + '工作簿1.xlsx')

df.head()

# 使用 sklearn 進行分層抽樣
from sklearn.model_selection import train_test_split

# data['TYPE']是在data中的某一個屬性列
X_train, X_test, y_train, y_test = train_test_split(df,df['TYPE'], test_size=0.2, stratify=df['TYPE']) # test_size 測試集占比

print(X_train.shape)
print(X_test.shape)
'''
(885, 4)
(222, 4)
'''


# 普通方法進行分層抽樣
test = pd.DataFrame()              # 划分出的test集合
train = pd.DataFrame()             # 剩余的train集合
tags = df['TYPE'].unique().tolist() # 按照該標簽進行等比例抽取

for tag in tags:
    # 隨機選取0.2的數據
    data = df[(df['TYPE'] == tag)]
    sample = data.sample(int(0.2*len(data)))
    sample_index = sample.index
    
    # 剩余數據
    all_index = data.index
    residue_index = all_index.difference(sample_index) # 去除sample之后剩余的數據
    residue = data.loc[residue_index]  # 這里要使用.loc而非.iloc
    
    # 保存
    test = pd.concat([test, sample], ignore_index=True)
    train = pd.concat([train, residue], ignore_index=True)

print(test.shape)
print(train.shape)
'''
(221, 4)
(886, 4)
'''