Appealing Risk data science models in banking: the buddybank project

In the last decade, has increased the attention on credit risk modelling. The primary model in credit scoring is the Logistic Regression because of its easy implementation and evaluation. In spite of this, the logistic regression is not the only statistical tool to extract knowledge from credit data. The main goal of this study is to develop different models starting from parametrical models like logistic regression to non-parametrical model like the Generalize Additive Model (GAM), using decision tree algorithm and lastly, the Generalized Extreme Value (GEV) model for rare events. The steps undertaken in the case study to perform and improving the knowledge process were variables selection, data preparation, data extraction and evaluation of results. The research shows how to improve the evaluation of credit worthiness and credit data quality.