Analysis Of Mobile Object Behavior Pattern Based On Location Information

As the rapid development of mobile devices and GPS positioning technology,mobile objects using location based service of location data is growing exponentially.A large number of location data contains abundant time information and space information.Through the study of mobile objects over time and space change rule,can be found that the behavior pattern of the mobile objects.It is of great significance for the object behavior prediction,personalized services recommendation,etc.Recently,most research on behavior pattern extraction based on GPS coordinates,can't reflect the intention of the movement,interests and hobbies of the mobile objects.In the other hand,because of the location data of high dimension characteristic,directly ha ndle real GPS coordinates leads to heavy work.In view of the shortcomings of the above behavior pattern mining of mobile object,this paper focuses on two problems: high dimension sparseness and timing.Aiming at the particularity of position data,this paper proposes a method of locating semantic discovery,mining potential regional functional characteristics under location coordinates,mapping high-dimensional location coordinate data to low-dimensional location semantic space.Based on the method of location semantic discovery,a group classification method based on location semantics and probability is proposed.The access probability of object to location semantic space is taken as their static features and applied to group classification.At the same time,the timing information included in the position data,the rules of moving object trajectory extraction are defined,and the constraints of three moving trajectories are established.A uniform and reasonable trajectory extraction mode is established for different time periods,different geographical distance a nd different location semantics.Digging the moving trajectory of a specific time slice of the mobile object,and detect its frequent movement as the object behavior pattern.The behavior pattern is applied to the group classification as the dynamic features of the mobile object.Based on the dynamic time warping(DTW)algorithm,the behavior pattern similarity calculation mechanism is established,and the maximum similarity between mobile objects is calculated as the classification basis of the mobile object behavior pattern.The validity of the static and dynamic features is verified by the results of the group classification.Experiments on the collected real location data sets show that the method of a group classification method based on location semantics and probability can accurately describe the location coordinates.Compared with other dimensionality reduction algorithms,it has higher performance.The group classification method based on the static feature of the mobile object has a good F-measure,which can effectively judge the mobile object type.The behavior pattern of the mobile object obtained by the behavior pattern extraction method conforms to the real daily movement pattern.The results of group classification based on the dynamic feature of mobile objects show that the dynamic feature of the object is more suitable for static feature of the object,and the F-measure value obtained in the group classification is better.