Interpretable Machine Learning – Increase Trust and Eliminate Bias
by Rob Chang, Fiona Meng, and Russell Rollow While today’s models have unprecedented levels of accuracy, this has come at the cost…
Education Student Loan Credit Modeling System
Student loan valuation challenger model.
Client: U.S Department of Education, Federal Student Aid
Challenge: Federal Student Aid (FSA) manages a $1.5 trillion portfolio of student loans, requiring a deep understanding of the portfolio composition, historical performance, and risk drivers. To do this, FSA needed to better understand its complex and disparate data sources, more easily track portfolio performance, and accurately forecast portfolio cash flows. They needed a challenger model to compare against the results of their in-house model.
FI Solution: FI inventoried and validated FSA’s data sources that were relevant to understanding risks to the portfolio. We assessed the quality of the data for accurate reporting and modeling purposes. We created Tableau dashboards to communicate portfolio composition and historical performance and trends such as defaults, prepayments, and recoveries. Finally, FI developed forecasting models to calculate a valuation of the portfolio under various scenarios.
FI Impact: FI provided FSA an independent valuation of its student loan portfolio on both a federal accounting basis and a fair value basis. The student loan modeling system we developed supports FSA’s future-state portfolio management capabilities and provides the Department with a suite of analytics for increased transparency. This project was time-sensitive and was completed in 5 months.
Related Content
