Multichoice Modeling: Exploring Decision Configuration Space under Uncertainty
Resumo indisponível. Consulte o paper original.
PAPER / ARXIV:2610.13199
Dian Prinsloo, Tanja Verster, Neill Smit
RESUMO
Credit risk modelling typically uses credit history, repayment behaviour, income, and outstanding debt. Although transactional data are sometimes incorporated through aggregated measures, behavioural information in raw transaction narratives remains largely unexplored. Using data from a large South African financial institution, this study applies NLP and regular expressions to transform unstructured transaction descriptions into explainable behavioural features via a novel categorisation algorithm. Incorporating these features into a credit risk scorecard resulted in a significant increase in Gini compared to a scorecard without them.
NO MESMO MAPA
Resumo indisponível. Consulte o paper original.
Resumo indisponível. Consulte o paper original.