A New Cluster Validity Index for Fuzzy Clustering Using Separation and Compactness

Abstract

Aiming at weak point of classical validity index based on Euclidean distance, we proposed a new validity index which based on Non-Euclidean Distance. First of all, we constructed compact index by using improved membership degree and then made use of anti-close degree to constructed separation index. Finally, a new validity index based on these was proposed. This method overcame the drawbacks of classical validity index based on Euclidean distance, which did not correctly reflect the relation of intra-class and inter-class. Experiments on different datasets show that new method has good performance in noise data environment. The capacity of new method is superior to others with regard to cope with overlap dataset. New method has more obvious superiority especially on high dimensional dataset.