Data Stream Clustering Algorithm Based on Grid Density
Lingjun Zou · Journal of Chinese Computer Systems · 2009
A real time data stream clustering algorithm named RTCS based on density analysis is presented.The algorithm consists of tow phases of online and offline processing.The online process receives the data from the fast data stream and dynamically calculates the densities of the high dimensional data and data grids,while the offline process forms the initial clusters and adjusts the clusters adaptively according to the changes on densities of data grids.The algorithm can distinguish the real isolated data according to the change of their densities and remove them without affecting the later clustering result.Experimental results show that RTCS can discover clusters with arbitrary shapes effectively.Compared with CluStream,RTCS has higher clustering quality and speed and is extendable for data sets of larger dimensions and sizes.