Point Of Interest Recommendation Method Based On User Behavior Trajectory Analysis

Under the background of the rapid development of Internet,intelligent mobile devices,location-based service software,GPS service and people's demand for location-based service,a large amount of location-based action data has been generated.These data based on spatiotemporal trajectories also greatly promoted the development of social networks.Effectively utilizing and mining the information contained in these spatial and temporal data is an urgent problem to be solved in real life,which can bring huge profits to enterprises in real life,but also increase the personalized service experience for users,and increase the stickiness of users in using software.There are also many new challenges and opportunities in mobile behavior modeling and behavior prediction.However,it remains a huge challenge to effectively model user check-in behavior on large data sets to make accurate recommendations.These challenges include issues of data sparsity,spatio-temporal heterogeneity,long-and short-term dependencies of sequences,and the impact of user relationships.Therefore,the prediction of the user's movement behavior trajectory and the recommendation of the next point of interest have also become a recent research hotspot.In view of these challenges above,this thesis carries out research work on relevant contents recommended for the next point of interest,and the main research results are as follows:According to the existing related theories of trajectory analysis,this thesis proposes a user behavior trajectory analysis method based on users' long-term and short-term interactive preferences.Through the comprehensive analysis of long-term user preferences and short-term user preferences from multiple perspectives,and the influence of time and space factors on user behavior trajectory is added.By means of vector embedding,the behavior trajectory characteristics of users are analyzed by using hierarchical attention mechanism,so as to prepare for the recommendation of points of interest.Ulteriorly,This thesis proposes a new method based on user behavior trajectory analysis for the recommendation of the next point of interest.Based on the LSTM model,this model integrates users' sequence information,users' long-term and short-term interaction preferences,activity trajectory and spatiotemporal information for the recommendation of interest points.In the end,this thesis experiments on real public datasets show that the proposed model is significantly better than the mainstream POI recommendation model and the latest sequential recommendation model.