Inferring course enrollment from partial data

Abstract

We study how to infer students’ course enrollment information from incomplete data. We use data collected from a leading technology company and use a novel extension of Factorization Machines that we call Weighted Feat2Vec. Our empirical evaluation suggests that we improve on popular methods, while training time is reduced by half (when using the same implementation language, and hardware).