B0192
Title: Statistically enhanced learning
Authors: Christophe Ley - University of Luxembourg (Luxembourg) [presenting]
Andreas Groll - Technical University Dortmund (Germany)
Florian Felice - University of Luxembourg (Luxembourg)
Stephane Bordas - University of Luxembourg (Luxembourg)
Abstract: Statistically enhanced learning (SEL) is a new approach to improving learning performance by preparing and augmenting a data set using statistical tools. A formal definition of SEL and a framework for understanding its different components are presented. SEL inherits from three different fields: learning, enhanced (data processing), and statistics. The framework identifies the intersections between these fields and defines different levels of SEL features. The levels range from proxies, which add new features to represent variables that cannot be observed, to MLE-based features, which extract information from available variables using more advanced statistical tools. Examples of how SEL can be applied to different learning problems are provided, such as football performance prediction. Researchers and practitioners are enabled to better understand and apply SEL to a wide range of learning problems.
