Optimized fuzzy clustering by fast search and find of density peaks

Clustering by fast search and find of density peaks (CFSFDP) is a new clustering method that can find the cluster centers, quickly. The core idea of CFSFDP is the depiction on cluster centers, which based on two assumptions: the cluster center is surrounded by points whose density does not exceed it, and is relatively more distant from other denser points. Based on these two assumptions, CFSFDP can select the cluster centers manually depending on decision graph. But, in the complex and diverse clustering tasks, selecting the cluster centers through the decision graph still has a great limitation. To solve this problem, a Fuzzy CFSFDP clustering algorithm has been proposed, which can select cluster centers through fuzzy rules, automatically. However, the performance of Fuzzy CFSFDP is usually affected by the cutoff distance d
c
. Moreover, the clustering of data with complex manifold structure distribution is not effective, resulting in incorrect clustering results. In this paper, we propose to optimize the Fuzzy CFSFDP algorithm using manifold distance and standard deviation based cutoff distance. We compared the Optimized Fuzzy CFSFDP with original Fuzzy CFSFDP on two synthetic datasets, clustering results verified the effectiveness of our method.