A0353
Title: Statistically enhanced learning for better predictions
Authors: Christophe Ley - University of Luxembourg (Luxembourg) [presenting]
Florian Felice - University of Luxembourg (Luxembourg)
Andreas Groll - Technical University Dortmund (Germany)
Abstract: Statistically enhanced learning (SEL) is presented, which is a general approach to improve any learning technique, be it statistical or machine learning, by adding highly informative covariates obtained as statistical estimates rather than directly observed. SEL works for any data (tabular, computer vision, text). The general idea is discussed, referring to existing feature extraction methods which actually can be shown to fall under the umbrella of SEL, and its performance is illustrated on both simulated and real data. In particular, it is shown how SEL allows improved predictions of soccer tournaments and discusses how it can be used in sports medicine for injury prevention.