如何固定scikit-learn评估器参数且兼容Pipeline、GridSearchCV等工具

问题根因 #

scikit-learn的get_params()方法依赖解析评估器__init__方法的显式参数签名收集参数，你之前自定义子类时仅使用**kwargs作为入参，没有显式声明父类的所有参数，导致get_params()无法识别参数，返回空字典。

该方案完全符合scikit-learn评估器开发规范，可强制锁定目标参数不被覆盖，完美适配Pipeline、GridSearchCV等所有工具。

实现代码 #

from sklearn.ensemble import RandomForestClassifier

class FiveTreesClassifier(RandomForestClassifier):
    # 显式声明所有父类原生参数，和父类__init__签名完全一致
    def __init__(
        self,
        *,
        bootstrap=True,
        ccp_alpha=0.0,
        class_weight=None,
        criterion="gini",
        max_depth=None,
        max_features="sqrt",
        max_leaf_nodes=None,
        max_samples=None,
        min_impurity_decrease=0.0,
        min_samples_leaf=1,
        min_samples_split=2,
        min_weight_fraction_leaf=0.0,
        n_estimators=100,
        n_jobs=None,
        oob_score=False,
        random_state=None,
        verbose=0,
        warm_start=False,
    ):
        # 直接传入固定的n_estimators到父类初始化，自动忽略用户传入的对应参数
        super().__init__(
            bootstrap=bootstrap,
            ccp_alpha=ccp_alpha,
            class_weight=class_weight,
            criterion=criterion,
            max_depth=max_depth,
            max_features=max_features,
            max_leaf_nodes=max_leaf_nodes,
            max_samples=max_samples,
            min_impurity_decrease=min_impurity_decrease,
            min_samples_leaf=min_samples_leaf,
            min_samples_split=min_samples_split,
            min_weight_fraction_leaf=min_weight_fraction_leaf,
            n_estimators=5, 
            n_jobs=n_jobs,
            oob_score=oob_score,
            random_state=random_state,
            verbose=verbose,
            warm_start=warm_start
        )

测试验证 #

fivetrees = FiveTreesClassifier()
randomforest = RandomForestClassifier(n_estimators=5)

# 两个断言均可正常通过
assert fivetrees.n_estimators == randomforest.n_estimators
assert fivetrees.get_params() == randomforest.get_params()

# 即使用户主动传入n_estimators也会被强制锁定为5
fivetrees2 = FiveTreesClassifier(n_estimators=100)
assert fivetrees2.n_estimators == 5

优缺点 #

• 优点：完全兼容所有scikit-learn工具，可强制锁定目标参数不被用户传入值、网格搜索等逻辑覆盖，其余参数可正常修改、调参
• 缺点：代码冗余度高，scikit-learn版本升级如果修改了父类的__init__参数列表，需要手动同步子类的参数签名

不需要定义新类，仅需一行代码即可固定参数，适配所有scikit-learn工具。

实现代码 #

from functools import partial
from sklearn.ensemble import RandomForestClassifier

FiveTreesClassifier = partial(RandomForestClassifier, n_estimators=5)

测试验证 #

fivetrees = FiveTreesClassifier()
randomforest = RandomForestClassifier(n_estimators=5)

# 两个断言均可正常通过
assert fivetrees.n_estimators == randomforest.n_estimators
assert fivetrees.get_params() == randomforest.get_params()

优缺点 #

• 优点：代码极简，无需维护参数签名，scikit-learn版本升级无兼容问题，其余参数可正常修改、调参
• 缺点：无法强制锁定参数，用户主动传入n_estimators时会覆盖固定值，例如FiveTreesClassifier(n_estimators=100)会得到100棵树的随机森林实例

内容的提问来源于stack exchange，提问作者Enrico Gandini