Point-of-Interest (POI) Recommendation Model Simulation using Social Relationships and Geographical Factors

Abstract

This research suggests a point-of-interest recommendation model that incorporates social interactions and some geographical parameters in order to address the issue of data sparsity in Point-of-Interest (POI) recommendation.The model first suggested a user similarity measurement method based on the shared sign-in and distance between users in the social relationship, and based on the collaborative filtering of users, it was possible to determine the degree to which the social relationship influenced the user's sign-in behavior.Second, divide the local activity area for each user, fully evaluate the check-in correlation between the nearby POIs that haven't been visited, and determine the extent to which the user's check-in behavior is influenced by the surrounding geography.Finally, a POI recommendation model that is more in line with the user's preference is constructed, which effectively solves the problem of data sparsity.It is based on the weighted matrix decomposition, the user's preference model and combined with the social relationship and local geographic factors considered above.Experiments on the Gowalla dataset show that the proposed SLGMF algorithm has good recommendation performance.