Choosing Aggregation Level for Forecasting and Fairness

Proposed a hierarchical forecasting scheme to forecast yearly college enrollment regarding race/ethnicity.

Created a bisecting hierarchical clustering algorithm to cluster time series and implemented the algorithm.

Performed hierarchical forecasting using statistical models.

Increased the forecast accuracy by 15% compared to the conventional enrollment forecasting scheme.

Provided a new method for institutions to modify their fairness presentation across groups by aggregating data.

Git Repo

Project Report