A Fast Incremental Clustering Algorithm Based on Grid and Density
Zhuo Chen, Liu Xiang-shuang, Zhuang Xiao-dong · 2007
In this paper, a fast incremental clustering algorithm based on grid and density called ICGD is implemented in order to realize the real time clustering of the dynamic data. The innovations of this algorithm is capturing the shape of data space by condensation points, and using grid-based and density-based clustering methods based on the theories of climbing hill algorithm and connectedness to cluster the data, guided by the difference data to implement incremental cluster. The algorithm has the ability of grid-based and density-based clustering methods' good features, overcoming the traditional grid-based clustering method's shortcoming of clustering quality debasement resulted by little or no consideration of data distribution when partitioning the grids. They can also decrease the number of region query and calculate in traditional density-based clustering method, which consequently reduces the I/O cost. Experimental results show that it can realize incremental clustering process effectively and accurately.