AE 12: Explaining models of student debt

Note

This application exercise is completed in class and submitted via a worksheet.

Discussion questions

1. Which features contribute the most to the random forest prediction for Cornell University? Which features contribute the least?

2. Interpret the Shapley values for the schools below. How do the predictions compare? Which features contribute the most to the predictions? How do the contributions differ between schools?

   • Harvard University
   • James Madison University

3. Which features are most/least important for predicting student debt? How does the model choice influence the relative importance of features?

4. How does the average faculty salary influence the predictions of the debt models?