Machine Learning in banking: Different approaches to credit risk in the first quarter of 2019

This paper analyzes recent literature searching for trends in machine learning applications for the banking industry, emphasizing credit risk. The research was conducted in comparative studies in relevant journals, resulting in 57 articles published between 1996 and 2019. In addition, information from the answers given in a questionnaire prepared for the employees of 12 banking institutions was used. Among them are Brazilian, Portuguese, and international banks (USA, Spain, France, United Kingdom and China).

The analysis was conducted using a dictionary of terms related to banking, credit risk, and machine learning.

This procedure allowed the identification of the relationships between terms and articles. The questionnaire results indicate that credit risk in banks is one of the main trends of use for machine learning, which confirms the survey guidelines. There is also a relevant interest in fraud detection. Several articles have focused on machine learning techniques, highlighting the use of regression models, neural networks, decision trees, and naive Bayes. In pointing out these current research results, this study also lists opportunities for future research.