A novel heterogeneous ensemble credit scoring model based on bstacking approach

Abstract

In recent years, credit scoring has become an efficient tool that allows financial institutions to differentiate their potential default borrowers. Accordingly, researchers have developed a myriad of approaches, including statistical and artificial intelligence techniques, to fulfill the task of credit scoring. Recent studies have shown that ensemble methods, which combine multiple algorithms that process different hypotheses to form a new hypothesis, generally outperform the other credit scoring approaches. In this paper, we propose a novel heterogeneous ensemble credit model that integrates the bagging algorithm with the stacking method. The proposed model differs from the extant ensemble credit models in three aspects, namely, pool generation, selection of base learners, and trainable fuser. Four popular evaluation metrics, including accuracy, area under the curve (AUC), AUC-H measure, and Brier score, are employed to measure the performance of alternative models. To confirm the efficiency of the proposed bstacking approach, a wide range of models, including individual classifiers, homogeneous ensemble model, and heterogeneous ensemble model, are introduced as benchmarks. We also provided a discussion on the accurate yet complex credit scoring model (e.g., bstacking) from a regulatory perspective.