Вы можете использовать sklearn's KFold :
import numpy as np
import pandas as pd
from sklearn.model_selection import KFold
# create dummy dataframe with 500 rows
features = np.random.randint(1, 100, 500)
labels = np.random.randint(1, 100, 500)
df = pd.DataFrame(data = {"X": features, "Y": labels})
kf = KFold(n_splits=10, random_state=42, shuffle=True) # Define the split - into 10 folds
kf.get_n_splits(df) # returns the number of splitting iterations in the cross-validator
print(kf)
for train_index, test_index in kf.split(df):
print("TRAIN:", train_index)
print("TEST:", test_index)
X_train, X_test = df.loc[train_index, "X"], df.loc[test_index, "X"]
y_train, y_test = df.loc[train_index, "Y"], df.loc[test_index, "Y"]
Пример взят здесь .