Comparison of the Effectiveness between FKCS and KNC

Abstract

This paper presented the effectiveness of clustering data – the results from when the Fuzzy Kernel Compactness and Separation (FKCS) were proposed to compare with the performance of the Kernel Noise Clustering (KNC). The FKCS and KNC had largely been developed through basic steps of the Fuzzy Compactness and Separation (FCS) and Noise Clustering (NC) respectively. Ring data sets were used to test the effectiveness of clustering data between the FKCS and the KNC. In doing so, the kernel function – the kernel gaussian function and the kernel polynomial function were used as testors. Along those processes, noise data was added in order to test noise-resistant capacity of Fuzzy Kernel C-Mean (FKCM), FKCS and KNC. The results of the tests provided quite knowledgable understanding that when trying out with the ring datasets, the FKCS with kernel gaussian function and kernel polynomial function could perform effectively clustering data. Moreover, the KNC was even more robust and could work out better with noise data than the FKCM or the FKCS could.