Machine Learning and new data sources for credit scoring.

Summary In this article, we propose a reflection on the contribution of Machine Learning techniques and New Data to credit risk modeling. Credit scoring was historically one of the first fields of application of Machine Learning techniques. Today, these techniques make it possible to exploit "new" data made available by the digitalization of customer relations and social networks. The combination of the emergence of new methodologies and new data has thus structurally modified the credit industry and favored the emergence of new players. First, we analyze the contribution of Machine Learning algorithms with a constant information set. We show that there are productivity gains linked to these new approaches but that the gains in credit risk prediction remain modest. Second, we evaluate the contribution of this "datadiversity", whether or not these new data are exploited by Machine Learning techniques. It turns out that some of these data reveal weak signals that significantly improve the quality of the assessment of borrowers' creditworthiness. At the micro level, these new approaches promote financial inclusion and access to credit for the most fragile borrowers. However, Machine Learning applied to these data can also lead to bias and discrimination.