如何在sklearn中修复此自定义转换器？

Question

I've written this simple custom transformer that fills na in specific columns with 0. When I fit_transform on my dataset, it does not fill nas in the specified columns. I've failed to see the problem in my code.

class CustomImputer(BaseEstimator, TransformerMixin): 
    def fit(self, X, y=None):
        return self
    def transform(self, X, y=None):
        for col in ('PavedDrive', 'GarageQual', 'GarageFinish', 'FireplaceQu', 'KitchenQual', 'CentralAir', 'HeatingQC', 'BsmtExposure', 'BsmtCond', 'BsmtQual', 'ExterCond', 'ExterQual', 'Street'):
            X[col].fillna(0)
        return X

I expected the returned dataframe to be one where the specified columns have filled nas with 0 however, I get a dataframe with the same null values.

Answer 1

You never assigned the fillna operation to a new variable. Using

X[col].fillna(0)

does not happen in-place. Instead use:

X.loc[:,col] = X[col].fillna(0)

Answer 2

One liner solution would be

cols = ['PavedDrive', 'GarageQual', 'GarageFinish', 'FireplaceQu', 'KitchenQual', 'CentralAir', 'HeatingQC', 'BsmtExposure', 'BsmtCond', 'BsmtQual', 'ExterCond', 'ExterQual', 'Street']

X.loc[:,cols] = X[cols].fillna(0)