Sophisticated Machine Learning (Ml) models have seen to increase predictive accuracy of linear regression models in the context of credit risk modelling. Nevertheless, linear regression models remain popular in the credit risk industry, because of the lack of transparency of Ml models. In this study, we propose a way to interpret, tuning and ex- tract default probabilities from Ml technologies in the context of credit risk. Using a sample of Italian Small and Medium sized Enterprises’ (Smes), we show how much and why Ml models increase predictions and precision of default probabilities.
