《Improved insensitive to input parameters trajectory clustering algorithm》.pdfVIP

Journal of Systems Engineering and Electronics Vol. 24, No. 5, October 2013, pp.852–861 Improved insensitive to input parameters trajectory clustering algorithm Jiashun Chen 1,2,* and Dechang Pi1 1. College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China; 2. College of Computer Science and Technology, Huaihai Institute of Technology, Lianyungang 222003, China Abstract: The existing trajectory clustering (TRACLUS) is sensi- in the process of trajectory cluster analysis, most of clus- tive to the input parameters ε and MinLns. The parameter value ter algorithms need users to input some parameters [7–11] is changed a little, but cluster results are entirely different. Aiming such as density threshold, distance threshold, number of at this vulnerability, a shielding parameters sensitivity trajectory cluster. Furthermore, cluster results are often sensitive to cluster (SPSTC) algorithm is proposed which is insensitive to the input parameters. Firstly, some definitions about the core distance these parameters. If the parameters are selected improp- and reachable distance of line segment are presented, and then erly, the quality of cluster results will be poor; mean- the algorithm generates cluster sorting according to the core dis- while, the selected parameters often depend on experience tance and reachable distance. Secondly, the reachable plots of or expert knowledge. In addition, the selected parameters line segment sets are constructed according to the cluster sor- are often very difficult, especially for high-dimensional ting and reachable distance. Thirdly, a parameterized sequence is data sets. It is also difficult to acquire good cluster re- extracted accordi