Rogelio Andrade Mancisidor, PhD Candidate UiT (University of Tromsø) & Santander Consumer Bank
How to use Deep Variational Auto Encoders in Segment-based credit risk assesments
It has been shown that segment-based credit scoring, in some cases, can improve the overall classification accuracy in the credit risk assesment for new customers applying for a loan product. Currently, Rogelio has developed a model using the flexibility of Deep Variational Auto Encoders to identify hidden segments, in a given customer portfolio, with significantly different risk profiles. Financial institutions can leverage this valuable segmentation to conduct segment-based credit risk assesment, i.e. build classification models on each segment.
Rogelio is a PhD candidate in Machine Learning at the Machine Learning Group at the University of Tromsø. His research focus is Deep Learning models for credit risk. Specifically, he works with Deep Variational Auto Encoders, which are a specific methodology in the generative models. The research project is funded by Santander Consumer Bank - Nordics. The focus of Rogelio’s research is on applications of different Machine Learning techniques in the field of Credit Risk Modeling using Santander Bank’s data. Credit Risk arises with the uncertainty of whether a client will pay back a granted loan. Machine Learning has shown to outperform traditional classification techniques in this field.