Assignment 5
MGMT 638: Data-Driven Investments: Equity
Rice University

Instructions

Submit a pdf to Canvas. This assignment is due by midnight on Wednesday, December 6.

Backtest at least two random forest models for your sector, one for each definition of the target. Other models could be backtested using different max_depth or a different number of years for training data or a different number of stocks or excluding or including large caps, etc. Your evaluation should contain the following:

  1. Is the strategy successful in identifying good stocks and bad stocks?
  2. Does it produce alpha relative to the market?
  3. Does it have significant factor betas and does it produce alpha relative to the factor model?
  4. What features does it seem to be using and how - that is, does it produce higher forecasts when the feature is larger or when the feature is smaller?