We present a new method to address the problem of personalized behavior classification and apply it to depression detection among college students. Inspired by the idea of collaborative filtering, our method is a type of memory-based learning algorithm. It leverages the similarities of mobile-sensed behavior features among individuals to calculate personalized relevance weights, which are used to impute missing data and generate intermediate predictions (e.g., whether the student would have depressive symptoms).

This project has undergraduate student research opportunities.