python - train_test_split 不拆分数据

Question

有一个dataframe，共有14列，最后一列是目标标签，整数值为0或1。

我已经定义了:

1. X = df.iloc[:,1:13] ---- 这由特征值组成
2. y = df.iloc[:,-1] ------ 这由相应的标签组成

两者的长度都是一样的，X是一个由13列组成的dataframe，形状是(159880, 13)，y是一个数组类型，形状是( 159880,)

但是当我在 X,y 上执行 train_test_split() 时，该函数无法正常工作。

下面是简单的代码:

X_train, y_train, X_test, y_test = train_test_split(X, y, random_state = 0)

拆分后，X_train 和 X_test 都具有形状 (119910,13)。 y_train 的形状为 (39970,13) 而 y_test 的形状为 (39970,)

这很奇怪，即使在定义了 test_size 参数之后，结果仍然保持不变。

请指教，可能出了什么问题。

import pandas as pd
import numpy as np
from sklearn.tree import DecisionTreeClassifier
from adspy_shared_utilities import plot_feature_importances
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression

def model():
    
    df = pd.read_csv('train.csv', encoding = 'ISO-8859-1')
    df = df[np.isfinite(df['compliance'])]
    df = df.fillna(0)
    df['compliance'] = df['compliance'].astype('int')
    df = df.drop(['grafitti_status', 'violation_street_number','violation_street_name','violator_name',
                  'inspector_name','mailing_address_str_name','mailing_address_str_number','payment_status',
                  'compliance_detail', 'collection_status','payment_date','disposition','violation_description',
                  'hearing_date','ticket_issued_date','mailing_address_str_name','city','state','country',
                  'violation_street_name','agency_name','violation_code'], axis=1)
    df['violation_zip_code'] = df['violation_zip_code'].replace(['ONTARIO, Canada',', Australia','M3C1L-7000'], 0)
    df['zip_code'] = df['zip_code'].replace(['ONTARIO, Canada',', Australia','M3C1L-7000'], 0)
    df['non_us_str_code'] = df['non_us_str_code'].replace(['ONTARIO, Canada',', Australia','M3C1L-7000'], 0)
    df['violation_zip_code'] = pd.to_numeric(df['violation_zip_code'], errors='coerce')
    df['zip_code'] = pd.to_numeric(df['zip_code'], errors='coerce')
    df['non_us_str_code'] = pd.to_numeric(df['non_us_str_code'], errors='coerce')
    #df.violation_zip_code = df.violation_zip_code.replace('-','', inplace=True)
    df['violation_zip_code'] = np.nan_to_num(df['violation_zip_code'])
    df['zip_code'] = np.nan_to_num(df['zip_code'])
    df['non_us_str_code'] = np.nan_to_num(df['non_us_str_code'])
    X = df.iloc[:,0:13]
    y = df.iloc[:,-1]
    X_train, y_train, X_test, y_test = train_test_split(X, y, random_state = 0)    
    print(y_train.shape)

Answer 1

你搞错了train_test_split的结果，应该是

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2,random_state=0)