Identifying the most Informative Variables for Decision-Making Problems - a Survey of Recent Approaches and Accompanying Problems

Abstract

We provide an overview of problems related to variable selection (also known as feature selection) techniques in decision-making problems based on machine learning with a particular emphasis on recent knowledge. Several popular methods are reviewed and assigned to a taxonomical context. Issues related to the generalization-versus-performance trade-off, inherent in currently used variable selection approaches, are addressed and illustrated on real-world examples.