Research And Implementation Of Personalized Location Recommendation Algorithm Based On Hybrid Trajectory Similarity

Location based social network(LBSN)is the combination of online social network and offline physical world by using users' check-in information.It provides users with a new platform integrating virtual world and physical world,sharing and recommending virtual world information in real world services.Location recommendation system can help users find interesting locations by mining user preferences,social relations and location geographic information,which saves time and energy and attracts more user groups for location providers.Location recommendation system is widely used in various app services,such as map travel and travel recommendation.However,the location recommendation system still faces many problems,such as the lack of the ability to provide personalized recommendation services accurately and quickly in massive data,the difficulty to face the problem of data sparsity and the unreasonable and unreachable recommendation location.These problems limit the development and promotion of location recommendation system.Therefore,in order to solve the above problems,this thesis proposes a personalized location recommendation algorithm based on hybrid trajectory similarity by comprehensively considering the user's interest preference,mobile characteristics,geographical influence and access behavior,the main research contents are as follows(1)Because the original trajectory data is dense and huge,this thesis extracts the original trajectory data as stop points and positions to reduce the amount of data and retain the user's position information.Based on the idea of word frequency statistics,the location semantics is perceived,and two kinds of trajectories in geographic space and semantic space are obtained,which solves the problem of complex and time-consuming computation in massive data.(2)In order to meet the personalized needs of users,a user personalized interest preference model is established based on the idea of hyperlink induced search algorithm.Aiming at the problem of data sparsity,a hybrid trajectory similarity evaluation scheme based on user personalized interest preference model is proposed to evaluate and mine the location from the friend data with similar mobile characteristics with the semantic distance,geographical distance,time information,semantic familiarity and location popularity of the trajectory being considered.(3)In the process of actual recommendation,there are unreasonable recommendation problems caused by the distance between recommended locations.In this thesis,the influence of geographical factors on user's access behavior is modeled as the influence of personalized distance distribution.Based on kernel density estimation,user's personalized geographical distribution model is established to simulate user's access behavior to different recommended locations.(4)Combined with the hybrid trajectory similarity evaluation scheme and user personalized geographic distribution model,the access behavior and mobile characteristics of users are simulated,the probability of users accessing alternative recommended locations is calculated,and the final location recommendation results are pushed to users after sorting.To sum up,this thesis builds the user's personalized interest preference model to meet the user's personalized needs;According to the similarity of mixed trajectories,similar friends are evaluated to solve the problem of data sparsity;The personalized geographic distribution model is established to ensure that the recommended location is reasonable and accessible.Experiments show that the personalized location recommendation scheme proposed in this thesis has better recommendation performance in the scene of massive data and data sparsity.The research results of this thesis will provide theoretical guidance and technical support for the intelligent development and wide applications of social networks and recommendation systems.