INNER CODE UNIT · Python

predictor_columns

AnotherSamWilson/miceforest · miceforest/imputation_kernel.py:264

        predictor_columns = sum(self.variable_schema.values(), [])
        self.predictor_columns = [
            col for col in data.columns if col in predictor_columns
        ]

        # Make sure all pandas categorical levels are used.
        rare_level_cols = []
        for col in self.modeled_categorical_columns:
            value_counts = data[col].value_counts(normalize=True)
            if np.any(value_counts < 0.002):
                rare_level_cols.append(col)
        if rare_level_cols:
            warn(
                f"{','.join(rare_level_cols)} have very rare categories, it is a good "
                "idea to group these, or set the min_data_in_leaf parameter to prevent "
                "lightgbm from outputting 0.0 probabilities."
            )

View source record →