from sklearn.model_selection import KFold, cross_val_scorefrom sklearn.datasets import load_irisfrom sklearn.ensemble import RandomForestClassifier# 加载数据data = load_iris()X, y =>

from sklearn.model_selection import LeaveOneOutfrom sklearn.datasets import load_irisfrom sklearn.ensemble import RandomForestClassifier# 加载数据data = load_iris()X, y =>

import numpy as npfrom sklearn.utils import resamplefrom sklearn.metrics import accuracy_scorefrom sklearn.ensemble import RandomForestClassifierfrom sklearn.datasets import load_iris# 加载数据data = load_iris()X, y =>

from sklearn.model_selection import GridSearchCV, cross_val_score, KFoldfrom sklearn.datasets import load_irisfrom sklearn.ensemble import RandomForestClassifier# 加载数据data = load_iris()X, y =>

from sklearn.model_selection import StratifiedKFold, cross_val_scorefrom sklearn.ensemble import RandomForestClassifierfrom sklearn.datasets import load_iris# 加载数据集data = load_iris()X, y =>

期间序列交叉验证是一种专门用于期间序列数据的模型评价方法。与传统交叉验证方法（如 K 折交叉验证）不同，期间序列交叉验证遵照期间的人造顺序，从过去到未来逐渐划分数据。这种方法实用于期间序列预测义务，能够防止将未来的数据走漏到模型训练环节中，从而保障模型的实在性能评价。

期间序列交叉验证的罕用方法

滚动预测法

在滚动预测法中，每次划分训练和测试集时，训练集的大小坚持不变，仅在期间上向前滚动。每次只经常使用最新的训练集数据来预测未来的测试集数据。适宜模拟模型实时滚动预测的状况。

优缺陷：

import numpy as npfrom sklearn.linear_model import LinearRegressionfrom sklearn.metrics import mean_squared_error# 创立期间序列数据X = np.arange(1, 101).reshape(-1, 1)# 1到100的期间序列y = X.flatten() + np.random.normal(scale=5, size=X.shape[0])# 参与噪声# 定义滚动窗口的大小window_size = 20# 每次训练的窗口大小# 初始化模型model = LinearRegression()# 存储预测结果和误差predictions = []errors = []# 滚动窗口法启动预测for i in range(len(X) - window_size - 1):# 定义训练集和测试集X_train = X[i : i + window_size]y_train = y[i : i + window_size]X_test = X[i + window_size : i + window_size + 1]y_test = y[i + window_size : i + window_size + 1]# 训练模型model.fit(X_train, y_train)# 启动预测y_pred = model.predict(X_test)# 记载预测和误差predictions.append(y_pred[0])mse = mean_squared_error(y_test, y_pred)errors.append(mse)print(f"Window {i+1}: Predicted = {y_pred[0]:.2f}, Actual = {y_test[0]:.2f}, MSE = {mse:.2f}")# 计算平均误差average_mse = np.mean(errors)print(f"\nAverage MSE over all windows: {average_mse:.2f}")

扩展窗口法

在扩展窗口法中，训练集会随着期间始终扩展，即每次训练时，训练集蕴含更常年间段的数据，而测试集仍为后续的一小部分数据。这种方法能让模型在预测时应用更多历史数据。

优缺陷

from sklearn.model_selection import TimeSeriesSplitfrom sklearn.linear_model import LinearRegressionfrom sklearn.metrics import mean_squared_errorimport numpy as np# 创立期间序列数据X = np.arange(1, 101).reshape(-1, 1)y = X.flatten() + np.random.normal(scale=5, size=X.shape[0])# 经常使用TimeSeriesSplit成功滚动预测法tscv = TimeSeriesSplit(n_splits=5)model = LinearRegression()for train_index, test_index in tscv.split(X):X_train, X_test = X[train_index], X[test_index]y_train, y_test = y[train_index], y[test_index]model.fit(X_train, y_train)y_pred = model.predict(X_test)print("Test MSE:", mean_squared_error(y_test, y_pred)