Deep Learning and Machine Learning Techniques for Credit Scoring: A Review

Credit scoring is one of the most important credit decision-making in banking institutions by collecting, analyzing, and classifying various credit elements and variables of customer financial data. Nowadays, there is an increase in research related to machine learning (ML) and deep learning models (DL) to improve accuracy. This has led to the emergence of various ML and DL methods as a core practice in the field of credit scoring using various datasets. The aim of this study is to provide in-depth insights on various ML and DL-based credit scoring techniques. For this purpose, articles published between 2018 and 2023 were systematically reviewed by formulating research questions, defining search terms, and filtering articles using predefined inclusion and exclusion criteria. In particular, the reported model type, dataset, key performance parameters, publication profile, and keywords were extracted, and then the results of the identified models were examined. Finally, the most important aspects of the DL and ML methods in credit scoring were discussed. It was noticed that the performance of ML and DL-based credit scoring models has generally been evaluated using accuracy and area under curve. It was also observed that the UCI datasets have been used as a benchmark in the development of advanced credit scoring algorithms. The study also shows that comparing the performance of DL and ML models for credit is difficult due to the heterogeneity of the reported performance metrics. Hybrid and ensemble model based credit scoring techniques are becoming more popular and are the most commonly used credit scoring model. Further, the gaps and future research directions were highlighted. This review is expected to serve as an up-to-date and comprehensive reference for interested researchers seeking to quickly understand the current progress in DL and ML methods for credit scoring.