Design And Implementation Of Personalized Recommendation System Based On Trajectory Mining In Mobile Environment

As the rapid development of information technology and the increasing information content,the problem of ?information overload? is getting worse.Recommendation systems have been widespread concerns from all works of life,because they can help people to find what they may be interested in.And many researchers have studied it in-depth and applied it to reality,achieving great achievements.However,traditional recommendation systems often take users' preference data on the project into account only,while ignoring users' trajectory data characteristics,when they do recommending work.The trajectory data is a serialized representation of the motion behavior of the moving object.By analyzing and digging the trajectory data,it can reflect the movement rule and behavior pattern of the moving object.It has a very important significance to improve the accuracy of personalized recommendation systems and users' satisfaction.So,this thesis designs and implements a personalized recommendation system based on trajectory mining in mobile environment.The thesis focuses on the following:(1)The trajectory data is pretreated by using anabnormaldetection algorithm based on time-space Hausdorff distance.Based on the TRAOD(Trajectory Outlier Detection)algorithm,the Hausdorff distance of the line segment is improved by considering the time characteristic of the trajectory data in combination with the road network traffic,so as to improve the accuracy and efficiency of the abnormal trajectory detection.Using this algorithm to remove the abnormaldata in the trajectory,reducing the anomaly in the trajectory clustering.(2)A spatio-temporal road-network aware trajectory clustering algorithm(ST-NEAT)based on road network constraints is used to cluster trajectory,finding users' movement and behavior patterns.The algorithm takes into account both temporal and spatial attributes of the trajectory data,and can acquire the trajectory information of the user at different time periods.It not only improves the quality of clustering,but also helps to provide users with different periods of targeted recommendation services.(3)With the support of the above theoretical technology,a personalized recommendation system is designed and implemented.The system provides users with friend recommendations,location recommendations and other auxiliary personalized recommendation functions.And it can be more accurate for users to recommend similar users and places which are more in line with their preferences and so on.