Point of interest recommendations based on the anchoring effect in location-based social network services

Abstract

A point of interest (POI) recommender system (RS) is one of the representative research areas based on the location-based social network (LBSN). Most POI RS studies utilized various implicit information or social information to improve recommendation accuracy. However, majority of these studies overlooked the importance of users’ initial check-in information. Users are affected by their first input data in online services, and this phenomenon is called the anchoring effect. In POI RSs, few studies have analyzed the association with the anchoring effect while other RS domains already verified this effect. In particular, a research area, including POI RS, that focuses on the importance of the initial input does not exist. In this paper, we propose a latent Dirichlet allocation (LDA) model based on the anchoring effect for POI RS. This model emphasizes the importance of initial check-in data and is called the anchor-LDA. Experimental results showed that the anchor-LDA outperformed existing LDA-based POI recommender algorithms. Furthermore, we validated the importance of initial check-in information on the LBSN.