| With the advance in location-acquisition and mobile computing techniques,a series of terminals represented by smart phones are becoming more and more popular in people's production and life.And,various location-based services have come into being,which has brought a great convenience to people.While people are enjoying the service,a large amount of location data are accompanied.These data imply the individuality and generality of the moving object,so that the researcher can analyze the law of it so as to further improve the human life and improve the productivity of the society.However,trajectory data mining in the society is still at the primary stage of exploration,and the effective utilization rate is not high.This paper focuses on the difference and universality of trajectory data,and on the basis of existing research,it further studies the outlier detection and clustering of trajectory.The main work of this paper is as follows:(1)Research on trajectory outlier detection based on isolation mechanism.Most of the existing trajectory outlier detection algorithms only take into account the space characteristics of the locus,and only can detect some static points of the location outlier.They are complex in parameter settings and lack practical application value.Therefore,from the perspective of multiple factors,an Isolation-Based Trajectory Outlier Detection(IBTOD)is proposed in this paper.First,the spatial feature and temporal feature is extracted from the sub-trajectories,and then the isolated forest model is employed to fuse the multi factor features to get the anomaly value.Finally,the anomaly value is analyzed by boxplot,and the anomaly threshold and the anomaly trajectory are obtained.Experiments on real trajectory data sets and comparison with the two results of artificial marking and classical TRAOD algorithm show that the algorithm is effective.(2)Research on trajectory clustering based on uncertainty.Most of the existing trajectory clustering algorithms do not take into account the uncertainty of trajectory points caused by acquisition errors,and lack of protection of user privacy.Therefore,based on the uncertainty of the trajectory point,an Uncertainty-Based Trajectory Clustering is proposed in this paper.First,Geohash technique is employed to discretize the sub-trajectories,and then the distance matrix is calculated through the improved editing distance.Finally,clusters of trajectories are obtained through the extended DBSCAN algorithm.Experiments on real animal trajectory and hurricane trajectory data sets show that the algorithm is effective.(3)Design and implementation of trajectory acquisition and analysis system.Most of the existing location-based service applications do not excavate the trajectory data deeply,and the academic achievements are difficult to combine with the engineering application.Therefore,based on the algorithm research,this paper designs and implements a trajectory data acquisition and analysis system,which can collect the trajectory data uploaded by the mobile terminal.After collating the collected data,analysts can call the algorithm interface in the system to analyze the trajectory data. |
