These issues result from the fact that small part of the PIE prediction is attributed the explainers, after all, are not the decision-making process to the interaction of features via a black-box themselves. Therefore, the second area is receiving much model, with the goal to boost the predictive performance attention in recent years, that focuses on building models while maintaining interpretability. As that are inherently interpretable, such as rule-based models, such, the interpretable model captures the main decision trees, linear models, case-based models, etc., contributions of features, and the black-box model which do not need external explainers. While interpretable attempts to complement the interpretable piece by models have been advanced to achieve very competitive performance capturing the "nuances" of feature interactions compared to black-box models in some examples as a refinement. We design an iterative training by extensively searching the model space, interpretable models algorithm to jointly train the two types of models.