Supervised Local Modeling for Interpretability