reciprocal_features.py

import numpy as np
import pandas as pd

from sklearn.base import BaseEstimator, TransformerMixin 

class ReciprocalFeatures(BaseEstimator, TransformerMixin):
    def __init__(self):
        self.input_features = None
        self.rename_lambda = \
            lambda input_feature: f"_inv_{input_feature}"
    
    def fit(self, X, y=None):
        assert self.input_features == None
        self.input_features = X.columns.tolist()
        
        self.input_features.extend(list(
            map(self.rename_lambda, self.input_features)
        ))
        
        return self
    
    def transform(self, X):
        assert self.input_features != None
        assert np.all(X != 0)
        
        reciprocal_X = 1 / X
        reciprocal_X.columns = map(
            self.rename_lambda, reciprocal_X.columns
        )
        
        return pd.concat([X, reciprocal_X], axis=1)
    
    def get_feature_names(self):
        assert self.input_features != None
        return self.input_features