An Improved High-Density Sub Trajectory Clustering Algorithm

TRACLUS algorithm based on partition-and-group framework could not be distinguished the optimal partitioning accurately when the migration of trajectory points on both sides of corridor middle line was greatly offset, and the algorithm was sensitive to the input parameters. According to above deficiency, an improved high-density sub-trajectory clustering algorithm (HTRACLUS_DL) is proposed under the practical application background of a traffic corridor identification. Initially, sub trajectories are divided based on the spatio-temporal characteristic similarity of trajectories. Furthermore, a sub-trajectory parallel boundary method is constructed, which has higher precision than the partitioning algorithm used in TRACLUS. Additionally, sub-trajectory clustering center neighborhoods possess local high density and surrounded by lower density sub trajectories. However, the different sub-trajectory clustering centers are heterogeneity. Finally, a new sub-trajectory clustering algorithm is robust to input parameters based on sub-trajectory entropy. Experimental results based on trajectory data of mobile phone user in two cities show that HTRACLUS_DL could be solved the deficiency of TRACLUS. At the same time, the method obtains better clustering result based on spatio-temporal characteristics of sub trajectory and does not depend on parameter selection. HTRACLUS_DL could be identified traffic corridor of urban group effectively.