期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：12
出版社：Journal of Theoretical and Applied
摘要：The Fuzzy Subtractive Clustering (Fsc) method is applied in many fields because it is able to produce optimal clusters without requiring initial information of many groups as well as on the k-mean method. Unfortunately, in the Fsc method, there is a radius parameter that has a vital role in generating optimal clusters. The magnitude of this radius parameter is hypothetical to be influenced by the variability of the covariance matrix of the dataset. This study investigates the magnitude of radius parameter that resulted in optimal clusters on three datasets with high variability (dataset1), moderate variability(dataset2), and low variability(dataset3) on covariance matrices. In the clustering process, the squash factor and accept ratio parameters are made in constant, while the radius parameter is the determined variable that leads to the optimal cluster achievement. Clustering results are said to be optimal based on two criteria: each cluster consists of at least 2 members, and the clustering produces the smallest Ctm value. The results of this study recommend that prior to clustering with Fsc, it should be calculated first covariance matrix based on the standardized dataset. If the covariance matrix has a high variability, the radius value used is close to 1, the moderate variability is a radius value of about 0.5, whereas the low variability is used near the 0 radius value.
