クロスバリデーション（交差検証）についてわかりやすく解説

import numpy as np
from sklearn.model_selection import cross_val_score, KFold
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error

# サンプルデータの生成
np.random.seed(42)
X = np.random.rand(100, 3)  # 100行3列の特徴行列
y = 2 * X[:, 0] + 3 * X[:, 1] + 4 * X[:, 2] + np.random.randn(100)  # 重回帰式にノイズを加えた目標値

# 5フォールドのクロスバリデーションを設定
num_folds = 5
kf = KFold(n_splits=num_folds, shuffle=True, random_state=42)

# 重回帰分析モデルを設定
model = LinearRegression()

# 各フォールドごとの平均変数の係数を保存するためのリストを初期化
coeffs_per_fold = []

# クロスバリデーションで性能評価を実行
for fold, (train_index, test_index) in enumerate(kf.split(X), start=1):
    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)
    mse = mean_squared_error(y_test, model.predict(X_test))
    coeffs_per_fold.append(model.coef_)
    
    print(f"Fold {fold}: Mean Squared Error = {mse}, Coefficients = {model.coef_}")

# 最終的に平均した変数の係数を計算
average_coeffs = np.mean(coeffs_per_fold, axis=0)
print(f"Average Coefficients across {num_folds} folds = {average_coeffs}")

# 平均MSEを計算して表示
average_mse = -np.mean(cross_val_score(model, X, y, cv=kf, scoring='neg_mean_squared_error'))
print(f"Average Mean Squared Error across {num_folds} folds = {average_mse}")

'''
出力結果
Fold 1: Mean Squared Error = 2.0173924806103, Coefficients = [2.10589767 2.92460599 4.68605925]
Fold 2: Mean Squared Error = 0.6224586755912234, Coefficients = [2.19706932 2.82318627 4.53878099]
Fold 3: Mean Squared Error = 0.681337087852153, Coefficients = [2.39154581 2.90624531 4.44354592]
Fold 4: Mean Squared Error = 0.44423681259417125, Coefficients = [2.29699543 2.92552169 4.61495879]
Fold 5: Mean Squared Error = 1.4773918920339761, Coefficients = [2.45886327 2.6589557  4.62285239]
Average Coefficients across 5 folds = [2.2900743  2.84770299 4.58123947]
Average Mean Squared Error across 5 folds = 1.0485633897363646
'''

import numpy as np
from sklearn.model_selection import cross_val_predict, LeaveOneOut
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error

# サンプルデータの生成
np.random.seed(42)
X = np.random.rand(100, 3)  # 100行3列の特徴行列
y = 2 * X[:, 0] + 3 * X[:, 1] + 4 * X[:, 2] + np.random.randn(100)  # 重回帰式にノイズを加えた目標値

# Leave-One-Out Cross-Validation を設定
loo = LeaveOneOut()

# 重回帰分析モデルを設定
model = LinearRegression()

# LOOCV で性能評価を実行
y_pred = cross_val_predict(model, X, y, cv=loo)

# 各変数の係数を保存するためのリストを初期化
coefs = [[] for _ in range(X.shape[1])]

# LOOCV内でモデルをトレーニングし、各変数の係数を保存
for train_index, _ in loo.split(X):
    model.fit(X[train_index], y[train_index])
    for i, coef in enumerate(model.coef_):
        coefs[i].append(coef)

# 各変数の平均係数を計算し出力
for i, coef_list in enumerate(coefs):
    mean_coef = np.mean(coef_list)
    print(f"Variable {i+1}: Mean Coefficient = {mean_coef}")

# 平均MSEを計算して表示
mse = mean_squared_error(y, y_pred)
print(f"Mean Squared Error = {mse}")

'''
出力結果
Variable 1: Mean Coefficient = 2.2780064787295164
Variable 2: Mean Coefficient = 2.842209463099327
Variable 3: Mean Coefficient = 4.570304776138533
Mean Squared Error = 0.9962216632720223
'''

import numpy as np
from sklearn.datasets import make_classification
from sklearn.model_selection import StratifiedKFold
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score

# 仮想のデータ生成
X, y = make_classification(n_samples=1000, n_features=20, n_classes=2, random_state=42)

# 層化k分割交差検証を設定
num_folds = 5
skf = StratifiedKFold(n_splits=num_folds, shuffle=True, random_state=42)

# ロジスティック回帰モデルを設定
model = LogisticRegression()

# 各フォールドでの性能評価を実行
fold_accuracy_scores = []
for fold, (train_index, test_index) in enumerate(skf.split(X, y), start=1):
    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)
    accuracy = accuracy_score(y_test, y_pred)
    fold_accuracy_scores.append(accuracy)
    
    print(f"Fold {fold}: Accuracy = {accuracy}")

# 平均精度を計算して表示
average_accuracy = np.mean(fold_accuracy_scores)
print(f"Average Accuracy across {num_folds} folds = {average_accuracy}")

'''
Fold 1: Accuracy = 0.87
Fold 2: Accuracy = 0.885
Fold 3: Accuracy = 0.87
Fold 4: Accuracy = 0.86
Fold 5: Accuracy = 0.86
Average Accuracy across 5 folds = 0.869
'''