A Fast Method of Coarse Density Clustering for Large Data Sets

Density clustering algorithms are usually inefficient. Moreover, most of the density clustering algorithms needs an uncertain parameter of c which indicates the expecting amount of clusters. It makes the clustering results randomized by the unreasonable choice of c. And some non-density clustering algorithms also need such a parameter to be a precondition. So the inefficiency and random results of density clustering algorithms become a bottleneck of efficient and precise clustering. A fast method of Coarse Density Clustering(CDC algorithm) is presented in this paper. Its purpose is to find out the amount of the nature density cores of a sample space. It uses grids with a density greater than zero as processing units. CDC algorithm is more efficient and can be used to confirm the uncertain parameter of c for other clustering methods.