Geographical and Overlapping Community Modeling Based on Business Circles for POI Recommendation

Abstract

Point-of-interest (POI) recommendation is a challenging task since check-in data is extremely sparse and the social relationships in traditional recommendation have a limited effect. To solve this challenge, we propose a new geographical model with social influence and user preference. More specifically, we firstly propose a business circle conception which is more suitable for the modern consumption pattern in an urban city in POI recommendation. Then we decompose the user-location matrix into two geographical latent factors and integrate them into our business circle framework. Besides, we incorporate the user preference as a regularization of matrix factorization framework into our model by means of aggregating overlapping interest communities of users via their check-ins categories. Extensive experiments are conducted on two real-world datasets and the experimental results demonstrate that our model outperforms other existing algorithms.