Abstract:

Kernel Fuzzy C-Mean clustering (KFCM) utilizes kernel function to transform the data into the high-dimensional space, and uses the membership between data points and cluster centers to cluster the datasets. It is widely used in various fields due to its efficiency and fast speed. However, there are two limitations: the KFCM is sensitive to the initial values of clustering centers and it cannot automatically determine the number of clusters. Aiming at the two issues, we present an adaptive fuzzy clustering algorithm based on local search. The kernel method is introduced to improve the data separability, an evaluation index based on kernel is constructed to determine the number of clusters and the local search using small sample datasets is designed to look for the optimal cluster center. Experimental results on artificial datasets and the UCI dataset validate the effectiveness of the proposed method.

Key words: fuzzy clustering, Fuzzy C-Mean, kernel method, local search