Machine Learning applied to credit analysis: a Systematic Literature Review

Machine Learning (ML) has been increasingly used in credit analysis, also known as credit scoring. However, the vast majority of articles focus on ML techniques and do not delve into what are the most relevant variables to define good and bad payers. The objective of this research is to identify published works that study the variables that define the customer as a default or not, as well as to identify what leads the consumer to take credit even though he / she does not have the resources to (re-)pay. To achieve the objective of this study, a systematic literature review was carried out. The combination of automatic searches resulted in 36,639 articles, of which 17 were relevant. The studies found about the credit score present similar rating analysis methods, and only the variables used in the models changed.