Using function approximation for personalized point-of-interest recommendation

Point-of-interest (POI) recommender system encourages users to share their locations and social experience through check-ins in online location-based social networks. A most recent algorithm for POI recommendation takes into account both the location relevance and diversity. The relevance measures users' personal preference while the diversity considers location categories. There exists a dilemma of weighting these two factors in the recommendation. The location diversity is weighted more when a user is new to a city and expects to explore the city in the new visit. In this paper, we propose a method to automatically adjust the weights according to user's personal preference. We focus on investigating a function between the number of location categories and a weight value for each user, where the Chebyshev polynomial approximation method using binary values is applied. We further improve the approximation by exploring similar behavior of users within a location category. We conduct experiments on five real-world datasets, and show that the new approach can make a good balance of weighting the two factors therefore providing better recommendation.