Challenges and issues in trajectory streams clustering upon a sliding-window model

The proliferation of location-acquisition devices and thriving development of social websites enable analyzing users' movement behaviors and detecting social events in dynamic trajectory streams. In this paper, we firstly analyze the challenges in trajectory stream clustering, and then depict a three-part framework to deal with this issue, that includes i) trajectory data pre-processing for higher quality, ii) online micro-clustering to summarize a large number of microclusters, and iii) offline macro-clustering to form the resulting clusters. Particularly, we present the in-cluster maintenance strategy for online clustering evolving trajectory streams over sliding windows. It can eliminate the obsolete data while adaptively maintaining the summary statistics for continuously arriving location data, and thus avoid performance degradation with minimal harm to result quality.