Clustering high dimensional data streams at multiple time granularities

In this paper, we extend our DGStream (dense grid-tree based data stream clustering) method which is developed recently [Yan Xiaolong, et al., 2007] and propose a new method DGMStream (dense grid-tree based multiple time granularity adaptable data stream clustering) to cluster dynamic data streams. In DGMStream, we incorporate the technique of tilted time window in DGStream to find clusters for data streams over multiple time granularities. Implementation results show that this method has a better cluster purity and scalability than other methods.