A Comparison of Hash-Based Methods for Trajectory Clustering

The development of location-acquisition technologies has led to the emergence of massive spatial trajectory data. Recently many researchers have focused on techniques related to processing, managing and mining trajectories to extract knowledge and predictions useful for various applications. One of the first steps in processing trajectories is clustering and classification. Hash-based methods have been used and showed to be successful in clustering the large trajectory datasets. In this paper, we specifically focus on methods based on two types of hash functions: Locality-Sensitive and Distance-Based hash functions and compare them in terms of accuracy and bucket size balance. Our results suggest that, in comparison to Distance- Based hashes, Locality-Sensitive hashes results in higher accuracy but not necessarily higher bucket balance.