CMStatistics 2022: Start Registration
View Submission - CFE
Title: Machine learning techniques in joint default assessment Authors:  Patrizia Semeraro - Politecnico di Torino (Italy) [presenting]
Elisa Luciano - University of Torino (Italy)
Margherita Doria - Credit Suisse (Italy)
Abstract: The aim is to study the consequences of capturing non-linear dependence among the covariates that drive the default of different obligors in the overall riskiness of their credit portfolio. Joint default modeling is, without loss of generality, the classical Bernoulli mixture model. Marginal and joint defaults depend on a set of covariates, common to all obligors. Linear and nonlinear dependence among covariates is captured by ML methods, while LR captures linear dependence only. We show through an application to credit card data that the ability of machine learning methods to capture nonlinear dependence among the covariates produces higher default correlation and, therefore, more conservative risk measures of the quantile type.