Research On Point-of-Interest Recommendation Algorithm Based On Co-occurrence Graph

With the development of Mobile Internet and Location-based Social Network(LBSN),the recommendation of Point-of-Interest(POI)has become a popular research direction in industry and academia.Although a personalized POI recommender system can significantly facilitate users' outdoor activities,it faces many challenging problems,such as the hardness to model user's POI decision-making process and the difficulty to address data sparsity.Almost all of existing works focus on the modeling of POI recommendation from the user's point of view.However,the sparsity of data affects the effect of this method.To address these problems and challenges,a POI recommendation algorithm based on co-occurrence graph,called CoG,is proposed in this paper.The similarity of POI and geographical influence is integrated into the CoG algorithm for POI recommendation.In term of similarity,CoG uses the user's check-in information to construct the co-occurrence graph between the POI and define the two similarities between the nodes in the co-occurrence graph.Then,the recommendation problem is converted into the search of similar nodes in the graph.Specially,the similarity of POI is first applied into the poi recommendation,which make algorithm more robust and explainable.In term of geographical influence,we explore Gaussian kernel density estimation to model user's check-in location that makes recommendation more personalized.To evaluate the prosed CoG,We conduct extensive experiments with many related baseline methods on two real-world data sets,Foursquare and Gowalla.The experimental results show the effectiveness of CoG.