Abstract

GPRec explicitly categorizes users into groups in a learnable manner andaligns them with corresponding group embeddings. We design the dual groupembedding space to offer a diverse perspective on group preferences bycontrasting positive and negative patterns. On the individual level, GPRecidentifies personal preferences from ID-like features and refines the obtainedindividual representations to be independent of group ones, thereby providing arobust complement to the group-level modeling. We also present variousstrategies for the flexible integration of GPRec into various DRS models.Rigorous testing of GPRec on three public datasets has demonstrated significantimprovements in recommendation quality.

 

Quick Read (beta)