Data Stream Clustering Algorithm Based on Active Grid Density
Jing Yang, Wenxin Zhu, Jianpei Zhang, Yue Yang · 2010
CluStream is a representative data stream cluster algorithm, however, it is not able to find arbitrary clusters in shape and deal with boundary points effectively. A new algorithm which is based on active grid density AGD-Stream is proposed to solve these problems in this paper. The algorithm adopts a density decaying technology to capture the dynamic of data stream, and extracts the boundary point of grid to delete. Besides, this paper uses activity concept to judge activity of the grid density, neglecting the inactive grid density and preserving the active gird density for clustering. Experiment shows that AGD-Stream can not only find arbitrary clusters in shape, but also deal with boundary points. Compared with CluStream, AGD-Stream is better in time complexity and accuracy.