Behavioral Characteristic Modeling Based On Location Aware

With the rapid development of science and technology,the GPS positioning devices and the wireless communication technology have been used more popular and the degree of application has been greatly improved,the mobile terminal equipment which based on the location is increasingly intelligent,people can easily track of moving objects and collect their travel data by this equipment.The data contains a large number of mobile behavior rule,it can provide decision support for the urban traffic planning,the advertising push,and the research of human behavior’s characteristics.Therefore,the mobile behavior feature modeling has important research significance.This paper make the trajectory data of moving objects as the research object,with the moving objects behavior feature modeling and the motion trajectory destination prediction as the main target.In,this paper,the research work mainly includes the following four aspects:1.In the stage of data pre-processing,based on the real mobile trajectory data,this paper puts forward two kinds of dynamic meshing method to realize the space structure expression of the mobile trajectory area,they are Q meshing method and KD-tree meshing method respectively.Both the two method can alleviate the problems of rough spatial structure and low degree trajectory semantic that uniform meshing method causes.2.In the stage of feature modeling,aim at the data sparseness problem which often appear in prediction algorithm based on Bayes formula,this paper by means of the method of trajectory division and reconstruction,combine the Bayesian network theory and the Markov model,put forward two kinds of method to statistical step length,M1 method based on grid sequence number,and M2 method based on the trajectory sequence number.This paper also put forward a method to establish hybrid mobile behavior of feature modeling to overcome the sparseness problem.3.In the phase of track prediction,OD pattern matching prediction algorithm is presented in this paper as the basis prediction algorithm,and the OMD hybrid pattern1.matching prediction algorithm is proposed,which joined the mid point of trajectory as the local feature.The OMD hybrid pattern matching prediction algorithm could improve the trajectory matching degree.2.In the phase of revise the prediction result,based on geographic information model predicted results,this paper combine with fine-grained track prediction results based on the road network model to modify the original prediction result,improve the prediction precision of the trajectory.The Matlab2014 development platform is used in designing the experiment.The ShenZhen real taxi mobile trajectory data as the original database and the experiment set up based on it.Through the phases of data pretreatment,spatial structure partition,hybrid trajectory feature model establishment and patterns matching prediction,the experiment predicts the destinations of the query trajectories and comparing with actual destinations which belong to the query trajectories.The experiment results show that the destinations prediction algorithm’s predict rate of the query trajectories can reach 94.6%,it basically achieved the prediction of trajectory.