Moving Object Trajectory Mining Algorithms Based On XML

The mining technique of moving objects trajectory data is one of the important researchbranches of moving objects databases, and information derived from trajectory data can beapplied to traffic control, climate forecast, ecological study and so on. The amount of trajectorydata is very large and the structure is complex, in addition, due to the numerous types of the datasource and the diverse kinds of the data model, it is difficult to achieve effective management oftrajectory data using conventional data model. So there is an urgent need to develop a new modelto manage the trajectory data. The tree structure model is more flexible than the traditional datamodel, and the XML which is based on the tree structure is more suitable for the management ofthe trajectory data of the moving objects. At the same time the current research focuses on thefield of spatial data mining, which is only concerned about the current information. Therefore, it isnecessary to study data mining techniques for trajectories of moving objects.This paper mainly focuses on nearest neighbor query algorithms and similar trajectoryextraction methods of moving objects. Firstly, a formal definition of temporal distance based onthe trajectory data is provided, and in order to decreasing the amount of computation time in theprocess of query, this paper uses a relatively simply approximate distance to replace accuratedistance. Distance query, instant nearest neighbor query and interval nearest neighbor query basedon temporal distance are implemented on relational model and tree structure. The query efficiencyof two models is also compared, and the experimental results show that tree structure has betterefficiency. An index embedded within the XML document is proposed, which draws the idea ofthe spatial R-tree index. The data records that being in a certain space and satisfying certainconditions are stored together, so that it would reduce invalid operation during traversal of theDOM tree, and improve the query efficiency. Secondly, four similar trajectory query methods areproposed, which are nearest neighbor method, the average distance method, the standard deviationmethod and the confidence interval method, and the experimental results show that the confidenceinterval method is the most accurate method of them. Finally, we realize the above algorithms inthe real data set, and it needs to do coordinate conversion and standardization before conductingthe experiment. And the experimental results further demonstrate the validity of the proposedalgorithms.