Ziyue’s proposal involves using ensemble machine learning (a.k.a. super learning) for modeling healthcare expenditures. His first project proposes a two-stage super learner for dealing with healthcare expenditures with zero inflation and heavy right tails. Using simulations and data from the Medical Expenditures Panel Survey and the Back Pain Outcomes using Longitudinal Data, Ziyue showed improvements over existing methods for modeling healthcare data.
In his second project, Ziyue studied a super learner based on a Huber risk function. Ziyue proved theoretical results pertaining to the estimator’s behavior and again demonstrated impressive benefits in simulations and real data analysis.