Design And Implementation Of A Point Of Interest Recommendation System Based On Spatio-temporal Clustering

With the rapid development of mobile Internet and information technology,a large amount of mobile object location information is derived from the extensive application of modern smart device.Point of interest recommendation service is widely concerned by researchers at home and abroad because it helps users to timely and effectively obtain the content of interest from massive information.However,the traditional point of interest recommendation system only considers the location preference information of the moving object,but ignores the dispersion of the trajectory data of the moving object and the low density of the information distribution,which makes the trajectory information extraction difficult and the accuracy of the location analysis function of the point of interest recommendation system poor.Based on the above,a point of interest recommendation system based on spatial and temporal clustering is designed and implemented in this thesis.By converting spatio-temporal information into trajectory data,adding scenario labels of trajectory,comparing historical point of interest of users and comparing similarity of movement rules and behavior patterns among users,a highly accurate interest point recommendation system is realized.The main contents of this thesis are as follows:(1)Research on trajectory data preprocessing by using the Hierarchical Clustering Algorithm based on the Trajectory Segmentation(HCTS).Through the basic data cleaning algorithm and the improvement of the hierarchical clustering algorithm,the preliminary clustering results are formed to eliminate the redundant points and abnormal points in the massive trajectory data,which provide effective trajectory data and parameters for the realization of the next clustering algorithm and improve the quality of spatio-temporal trajectory data clustering.(2)Research on trajectory data clustering analysis by using the Clustering Algorithm Combining Multi-label Selection in Spatio-Temporal Trajectories(MS-SMoT).MSSMoT algorithm combines the time attribute and space attribute of the spatio-temporal trajectory to obtain the geographic location information of the moving object in different time periods.And then it adaptively adjusts the clustering radius size of clustering and accurately extracts the Point of Interest(POI)of the moving object in a single trajectory.This algorithm improves the accuracy of clustering algorithm and the validity of user location similarity measurement,and provides basic data for the point of interest recommendation system based on spatio-temporal clustering.(3)Design and implement of the point of interest recommendation system based on spatio-temporal clustering.The Point of Interest Recommendation Algorithm Based on Similarity Fusion(SFR)was designed and the linear Interest Point Recommendation system framework was constructed by combining location similarity and cosine similarity between users.The point of interest recommendation system is designed and implemented to provide users with more accurate and applicable interest points and recommendation contents of other auxiliary information,ensuring the adaptability and stability of the system.Through the functional test and performance test of the point of interest recommendation system,it is verified that the point of interest recommendation system in this thesis can provide users with location information in line with their interest preferences,and has high practicability in complex scenarios.