Graduation Prediction at GMU

In the summer of 2019 I worked under Dr. Huzefa Rangwala to predict four year graduation at George Mason University. These predictions will ultimately be used for academic publication and might eventually be extended help administrators know when to intervene to help students who might be struggling. I was particularly interested in the fairness aspect of this research – in the event that our algorithms are used by administrators I want to ensure that they don’t compound disparate impacts in some way. In order to accomplish this we had to discuss whether it would be more or less advantageous to be predicted as a potential dropout in the context that the algorithm would be used, and introduce unique metrics to measure how successfully the algorithm predicted dropouts of certain demographics versus others.