- Cowden, Chad [Browse]
- Senior thesis
- Fabozzi, Frank J. [Browse]
- Princeton University. Department of Operations Research and Financial Engineering [Browse]
- Princeton University. Program in Applications of Computing [Browse]
- Class year
- Summary note
- Loan default prediction is an essential component of the investment process of commercial real estate mortgage-backed securities. By looking at default projections for individual properties, investors can find relative value in different CMBS bonds. Investors in the CMBS loan world are fond of using stochastic default models in order to project default rates of properties, however an understudied approach to CMBS loan default prediction is the use of a binary classification model. This study focuses on the use of Support Vector Machines (SVM) to predict defaults for specific commercial real estate property data on Trepp’s bond database. By comparing the performance of SVM with that of logistic regression, another common binary classifier, the results of this study show that SVM does better to predict defaults on individual commercial properties. Moving forward with further research CMBS investors may consider methods of SVM prediction models, due to their better predictive performance, to find relative value in the underlying property assets in commercial mortgage backed security bonds.