Python中决策树的剪枝方法及其实现

# 定义后剪枝函数
def post_pruning(decision_tree, X_train, y_train, X_val, y_val):
    if decision_tree is None:
        return None
    if isinstance(decision_tree, Leaf):
        return decision_tree
    left_subtree = decision_tree.left
    right_subtree = decision_tree.right
    # 递归查找左右子树
    left_subtree = post_pruning(left_subtree, X_train, y_train, X_val, y_val)
    right_subtree = post_pruning(right_subtree, X_train, y_train, X_val, y_val)
    # 如果左右子树都是叶子节点，进行剪枝操作
    if isinstance(left_subtree, Leaf) and isinstance(right_subtree, Leaf):
        clf = DecisionTreeClassifier()
        clf.fit(X_train, y_train)
        acc_before_pruning = accuracy_score(y_val, clf.predict(X_val))
        merged_subtree = Leaf(left_subtree.most_common_class())
        acc_after_pruning = accuracy_score(y_val, predict(most_common_class(X_val, merged_subtree)))
        if acc_after_pruning > acc_before_pruning:
            return merged_subtree
    return decision_tree