Dempster-Shafer Theory in Recommender Systems: A Survey

Abstract

Due to the limitations associated with the use of a single type of data during the recommendation process, recent research has focused on developing new fusion-based recommenders that make use of multiple heterogeneous sources of information to provide more accurate suggestions. However, the realistic and flexible methods available to users for expressing their preferences for products and services inherently generate uncertain, imperfect, and ambiguous data that feed recommenders and thus affect their accuracy. As a result, Recommender Systems (RS) make significant use of soft mathematical tools to deal with uncertainty. Among these tools is Dempster-Shafer Theory (DST), which has been shown to be effective at dealing with the inherent uncertainty in numerous applications. This article provides a survey of the use of DST in the RS field. Thus, after a brief introduction to recommender systems and the DST, this survey discusses recent DST applications in RS. It introduces a new taxonomy that encompasses the primary application context for DST-based RS solutions, as well as a comprehensive multi-criteria analysis of the peer-reviewed papers. The resulting comparisons are analyzed to draw conclusions, identify current study limitations, and define future research directions. This survey serves as a valuable resource for the entire research community that is interested in recommender systems and DST.