Analysis Of Spatial And Temporal Data Based On Trajectory

With the rapid development of positioning technology and mobile Internet technology,we are able to get more and more mobile object position information;a variety of positioning equipment into the continuous miniaturization and popularization provides us obtain location information technical support;spatio-temporal data in recent years produces faster and faster,types becoming more and more;excavated from the mass of the space-time data to extract useful information draw more and more attention of researchers.Cluster analysis the researchers found that data mining can be a good dig out hidden in the complex and vast amounts of knowledge and information spatiotemporal data,this paper studied the temporal trajectory data clustering algorithms and data through time and space to build a user similarity analysis model.According to the article to improve the characteristics of the trajectory data of the traditional trajectory similarity measure from three angles and similarity symmetry full consideration to the characteristics of the trajectory data,and proposed a clustering algorithm on the basis of classical cluster centers by increasing localization ways to improve the accuracy of the traditional density clustering algorithm.The algorithm introduces the concept of cut off distance and the local density of trajectory,by selecting the range of areas in o cut off distance from the radius of the point of maximum local density as the cluster center,so close to the cluster center began clustering,clustering is more compact,reasonable,significantly enhance the clustering effect.To track users by collecting the information,the paper design a similarity analysis model based on user points of interest and region of interest.Firstly temporal clustering extracted interest points to obtain the set of points of interest for all users by setting the mean point of interest after a number of density levels in different points of interest on the cluster size to obtain the area of interest,there is a region of interest user's point of interest is considered to have similar patterns of behavior,as described above based on the idea to build a region of interest based on points of interest and user similarity matrix.This method takes into account overall factors of time and space,and improve the reliability of the algorithm by multi-granularity layered approach.In experiments by the hurricane and the user's GPS data sets show that the proposed clustering algorithm Clustering enables more compact and reasonable,similarity analysis model can better analyze user behavior similarity.