Research And Application Of A Point Of Interest Recommendation System Combining Time And Geographical Factors

With the popularity of location-based social networks,recommendation of points of interest has become an important task,which needs to learn users' preferences and mobile patterns to recommend points of interest.Previous studies have shown that it is necessary to integrate contextual information such as geography and time to improve the point of interest recommendation system.However,the existing approach is to model geographical impacts based on the physical distance between the points of interest and the user,while ignoring the impact of time factors.This paper studies user movement patterns and finds that user check-ins occur around several centers,depending on their current time state.The main innovations and improvements of this paper are as follows :(1)a new context model is proposed,which comprehensively considers geographical factors and time information;(2)a spatiotemporal activity center model is proposed to consider the user-centered behavior in different time states.(3)a static and time matrix model is proposed to study users' preferences and behaviors in static and time.In the static matrix decomposition,the model is trained on the whole user-interest matrix,while in the time matrix decomposition,different user-interest matrix training models are used in each time.(4)This paper also realizes the visual joint time factor and geographical factor of interest recommendation system.In order to prove the effectiveness of the method proposed in this paper,multiple experiments on two data sets,Gowalla and Foursquare,show that the model presented in this paper has statistically significant performance improvement compared with multiple comparison algorithms.At the same time,this paper also proves the effectiveness of obtaining geographic and temporal information to model user activity centers,and the importance of joint modeling of user activity centers.