python - How to fix Score function for it to work with GridsearchCV, TypeError: score() takes 2 positional arguments but 3 were given

Question

I'm trying to make my own BaseEstimator class to use it as part of a pipeline but i can't make it work. Here's what I get: TypeError: score() takes 2 positional arguments but 3 were given

I’ve already tried to add self argument to the score function, but it doesn’t work: i have a ValueError instead. When i make a separate model and then use that function separately, everything is just fine

class MyRegressor(BaseEstimator):

    def __init__(self, regressor_type: str = 'SGDRegressor'):
        """
        
        """
        self.regressor_type = regressor_type
        

    def fit(self, X, y):
        if self.regressor_type == 'SGDRegressor':
            self.regressor_ = SGDRegressor()
        elif self.regressor_type == 'RandomForestRegressor':
            self.regressor_ = RandomForestRegressor()
        elif self.regressor_type == 'LinearRegression':
            self.regressor_ = LinearRegression()
        elif self.regressor_type == 'CatBoostRegressor':
            self.regressor_ = CatBoostRegressor()
        elif self.regressor_type == 'XGBRegressor':
            self.regressor_ = XGBRegressor()
        else:
            raise ValueError('Unknown regressor type.')

        self.regressor_.fit(X, y)
        return self

    def predict(self, X):
        y_pred = self.regressor_.predict(X)
        return y_pred
      
    def score(y, y_pred):
        smape = sum(abs(y - y_pred) / (abs(y) + abs(y_pred)) / 2) * 100 / len(y)    
        return self.estimator.smape(y, y_pred) 

pipe = Pipeline([('scaler', StandardScaler()), ('MyRegressor', MyRegressor())])
pipe.fit(X_train, y_train_final)

params = {
    
    'MyRegressor__regressor_type': ['SGDRegressor', 'RandomForestRegressor', 'LinearRegression', 'CatBoostRegressor', 'XGBRegressor']
}


search = GridSearchCV(pipe , params, n_jobs=-1, cv=5)
search.fit(X_train, y_train_final)

print('Best model:
', search.best_params_)

Please help me to solve this issue. Thank you in advance!

question from:https://stackoverflow.com/questions/65940709/how-to-fix-score-function-for-it-to-work-with-gridsearchcv-typeerror-score-t

Answer 1

The expected signature of such a scoring function is (self, X, y). You should then use self.predict(X) to produce your y_pred inside the scoring function, then calculate the rest of the score as you have.