Abstract
Clustering by fast search and find of density peaks (CFSFDP) is a novel algorithm that efficiently discovers the centers of clusters by finding the density peaks. The accuracy of CFSFDP depends on the accurate estimation of densities for a given dataset and also on the selection of the cutoff distance (dc). Mainly, dc is used to calculate the density of each data point and to identify the border points in the clusters. CFSFDP necessitates using different methods for estimating the densities of different datasets and the estimation of dc largely depends on subjective experience. To overcome the limitations of CFSFDP, this paper presents a method for CFSFDP via heat diffusion (CFSFDP-HD). CFSFDP-HD proposes a nonparametric method for estimating the probability distribution of a given dataset. Based on heat diffusion in an infinite domain, this method accounts for both selection of the cutoff distance and boundary correction of the kernel density estimation. Experimental results on standard clustering benchmark datasets validate the robustness and effectiveness of the proposed approach over CFSFDP, AP, mean shift, and K-means methods.