A Neighborhood-Based Trajectory Clustering Algorithm
Yunxin Tao, Dechang Pi · 2008
Existing trajectory clustering algorithm TRACLUS uses global parameters, it can not distinguish small, close, and dense trajectory clusters from large and sparse trajectory clusters. Moreover, TRACLUS needs two input parameters and is sensitive to input parameters. To avoid the shortcomings of TRACLUS, a neighborhood-based trajectory clustering algorithm named NBTC is proposed based on the improved framework. Our key insight is that neighborhood-based local density is quite different from the absolute global density used in TRACLUS. NBTC keeps the efficient of TRACLUS and needs only one input parameter. Experimental results demonstrate that NBTC can discover trajectory clusters in arbitrary shape and different densities trajectory database effectively.