Abstract

Based on Dirichlet process, a Bayesian Model Averaging (BMA) method is proposed in this paper to predict residential mortgage default risk at aggregate level. This ensemble learning algorithm integrates estimates of the default risk provided by two machine learning and one classical forecasting models. Empirical analysis using regional mortgage default data shows that the Dirichlet BMA model performs better than other methods not only in fitting the data given, but also in providing out-of-sample predictions.

This work is licensed under a Creative Commons Attribution 4.0 License.