Survey on Popularity Based Recommendation

Abstract

Currently, popularity based recommendation has become a research hotspot. The use of popularity considerably improves the recommendation effects, while the Matthew effect caused by popularity bias has also garnered extensive attention among researchers. Some researchers consider combining both aspects to produce hybrid popularity based recommendation. Adopting the concept of popularity, a unified representation of popularity, popularity bias, and hybrid popularity is provided in this paper. Firstly, the background of popularity in the field of recommendation is introduced. Then, based on different perspectives, a comprehensive survey on popularity-enhanced recommendation methods, popularity debias recommendation methods, and hybrid popularity based recommendation methods is provided. Each type of method is further subdivided in specific subtasks of modeling or concrete strategies. The representative models of each method are introduced and analyzed, and their advantages and limitations are evaluated. The mechanisms and applicable scenarios of each method are also summarized in detail. Furthermore, the commonly used datasets, performance evaluation indicators and baseline are introduced. A comparative analysis of the representative methods performance is also listed. Finally, some opinions on the trends of popularity based recommendation are presented. An outlook on the technical difficulties and hotspots for future development from multiple perspectives is analyzed and predicted.

Keywords

• popularity; popularity bias; hybrid popularity; popularity based recommendation