Date of Submission
Campus Only Senior Thesis
Bachelor of Arts
This study evaluates the performance of linear model trees to forecast recovery rates of defaulted bonds. The linear model trees are built based on regression trees, with a linear regression model in each leaf. I use bond characteristics, firm characteristics, industry indicators, and macroeconomic indicators as explanatory variables. The relevance of explanatory variables is assessed using the Mutual Information Feature Selection method. The results show that the linear model trees present better out-of-sample forecasts of recovery rates in comparison with some other widely-used models.
lu, yiping, "Predicting Recovery Rates using Machine Learning Algorithms: the Relative Usefulness of Alternative Methods" (2020). CMC Senior Theses. 2389.
This thesis is restricted to the Claremont Colleges current faculty, students, and staff.