Towards comprehensive profile aggregation methods for group recommendation based on the latent factor model

Abstract

The aggregation of group members’ profiles is an extremely important step in group recommender systems, as it represents the whole group as a single virtual user, and is the input to traditional recommendation techniques. This paper focuses on proposing methods of aggregating profiles of group members. The highlight of the proposed methods lies in incorporating the latent factor matrices formed in the latent factor recommendation model into the profile aggregation for group recommendation. As a result, the process of the profile aggregation is ensured to more fully reflect the interests of the group members. Experimental results show that profile aggregation, enriched by latent factor matrices, can improve group recommendation performance in terms of F1-score and Normalized Discounted Cumulative Gain Measure.