Tra-DBScan: A Algorithm of Clustering Trajectories
Accompany with fast development of location technology, more and more trajectories datasets are collected on the real applications. So it is something of value in the theory and applied research to mine the clusters from these datasets. In this paper, a trajectory clustering algorithm, called Density-Based Spatial Clustering of Application with noise (Tra-DBSCAN for short), based on DBSCAN that is a classic clustering algorithm. In this framework, each trajectory firstly partitions into sub-trajectories as clustering object, and then line hausdorff distance is used to measure the distance between two sub-trajectories. Next, DBSCAN is introduced to cluster sub-trajectory to form cluster area, and then connecting different moments of clustering area is regarded as trajectory movement patterns. Finally, the experimental results show our framework’s effective.
Dongye Sun, Wen-Pei Sung and Ran Chen
L. X. Liu et al., "Tra-DBScan: A Algorithm of Clustering Trajectories", Applied Mechanics and Materials, Vols. 121-126, pp. 4875-4879, 2012