Clustering moving object trajectories using coresets

Abstract

Given a set of moving objects, we show the applicability of using coresets to perform fast approximate clustering. A trajectory coreset is simply a small weighted set of the trajectory segments that approximates the original trajectory. In this paper, we present an efficient algorithm that integrates coreset approximation with k-means clustering. Our methodology to build the trajectory coreset depends on using the trajectory segments movement direction. Using the movement direction feature of the segments we basically select the most influential segments to contribute in our coreset. The main strength of the algorithm is that it can quickly determine a clustering of a dataset for any number of clusters. In addition, to measure the quality of the resulting clustering, we use the silhouette coefficient. Finally, we present experimental results that show the efficiency of our proposed algorithm. © 2011 Springer-Verlag Berlin Heidelberg.