A Successive Points-of-interest Recommendation Algorithm For LBSN

With the rapid development of Internet,global technology and the continuous popularization of intelligent mobile communication devices,Location-Based Social Networks(LBSN)have penetrated into people's lives.With the help of portable terminal Internet devices,people can use the geographic information and social attributes in LBSN to define the geographical location preference of users' access Points-of-Interest(POIs),and arrange and adjust their travel plans in time.Due to the geometric growth of the number of users and check-in points in the current social network,massive data information appears in the social network.It has become a key research direction based on location recommendation to accurately mine users' preferred locations from such large scale check-in data and push them to the users accurately.Taking LBSN as the research background,this paper explores and analyzes the historical trajectories,personal preference and social factors of users.At the same time,considering the important role of geographical location clustering of Points-of-Interest and the correlation between POIs,this paper studies the algorithms of POIs recommendation and successive POIs recommendation.For the recommendation of POIs,a learning algorithm framework called Rank Friendship Bayesian Personalized Ranking(Rank-FBPR)is proposed.First of all,in order to make better use of the user's implicit feedback information,the user's personal preference function is obtained by using the Bayesian personalized ranking matrix decomposition framework FBPR based on social networks.Then use the Chinese Restaurant Process(CRP)to cluster POIs,and add the POIs aggregation information into the framework to get candidate POIs.Finally,using the personal preference function and power law distribution to predict the scores of candidate POIs,the sorted list of POIs is generated in descending order of the scores,and the list is recommended to the target user.The experiments verify the effectiveness of framework from both cold start and non-cold start.The experimental results on the Foursquare and the Yelp datasets show that Rank-FBPR has higher recommendation accuracy and recall than other comparison models,and can adapt to cold start problems.For the recommendation of successive POIs,a Context Geographical-Temporal-Social Awareness Hierarchical Attention Network(CGMS-HAN)is proposed.The model comprehensively considers the influence of user's trajectory sequences,time information,social relationship and geographical information of POIs,extracts context information from user's check-in histories,then designs Geographical-Temporal-Social attention network and a common attention network to learn dynamic user preferences.Finally reconstructs the objective function and learns model parameters through a negative sampling algorithm.The experimental results on Foursquare and Yelp real datasets show that CGTS-HAN are better than the comparison models in all four evaluation indicators under the same parameter setting,which confirms the effectiveness and superiority of CGTS-HAN.