Research On Moving Trajectory Data Analysis Method For Traffic Volume Overload Prediction

With the rapid development of economy,the urban traffic congestion has becoming an urgent problem in our country.The traffic volume overload prediction can provide an effective method for solving the problem.People always tend to find a convenient travel path and reach their destination smoothly.Usually,people are more familiar with the traffic situation in the region they live in.So,they can choose a proper path to avoid congestion,after the preliminary prediction of traffic volume.To take the traffic volume overload issue into account,this paper launches a research.By the means of trajectory acquisition,we can collect a large number of trajectory data of traffic moving objects.Then we research the methods of trajectory analysis,to predict the overload of traffic volume.So that we can help people to know about the urban traffic situation,and avoid the urban traffic congestion effectively.First of all,according to the analysis about a large amount of users’ trajectory data,a method(PRED)to analysis trajectory pattern dynamically and predict location is proposed.The first step is using the improved pattern mining model to extract trajectory frequent pattern(named T-pattern).Then we put forward the DPTConstruct_Update algorithm to design a data structure—DPT(Dynamic Pattern Tree),which contains spatio-temporal information and can store as well as query the trajectory frequency patterns of moving objects.In addition,the Prediction algorithm is presented to calculate the optimal matching degree and obtain the predicted location of moving objects’ trajectories.Secondly,a kind of trajectory clustering method(HMM-Cluster)based on the HMM model is put forward.With the extraction of trajectories’ feature points,we can decrease the trajectory data volume,as well as save the cost of storage.We train a HMM model for each trajectory.And then Sim-HMM algorithm is proposed to gain a trajectory affinity matrix,and then we propose the HMM-Cluster algorithm,which can contribute to aggregate similarity trajectory effectively,and find the overload of traffic volume.A series of contrast experiments based on actual data,prove that the PRED and HMM-Cluster methods have good effects.With the methods,we can obtain moving objects’patterns as well as predict traffic volumes effectively and conveniently.They have great application significances in our real life.