Abstract
Clustering is a well known technique in many areas of science and has been applied in such diverse fields such as botany, zoology, psychology, and market research. Cluster analysis is a multivariate analysis technique that seeks to organize objects described by a number of attributes or variables into relatively homogeneous groups, or ‘clusters'’. This paper presents a new developed clustering method for multivariate data. First, I describe four widely used clustering methods, SINLINK, K-means, CLASS, and Forgy's method. Next, the construct of the algorithm is discussed. Finally, comparisons of the four methods and the new method are described. From these comparisons, new method and CLASS are more suitable for the large data than the others. The best performance in the comparisons of the five methods is the proposed method.
