文章基本信息

标题：Performance Measure of Hard c-means,Fuzzy c-means and Alternative c-means Algorithms
本地全文：下载
作者：Binoda Nand Prasad ; Mohit Rathore ; Geeta Gupta 等
期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2016
卷号：7
期号：2
页码：878-883
出版社：TechScience Publications
摘要：Clustering analysis can be used to classify theobjects into subsets with similar attributes. The mainobjective of clustering techniques is to group the data points ina multi-attribute dataset such that the similarities aremaximized within the same cluster and minimized betweendifferent clusters. It is a branch in multivariate analysis andan unsupervised learning in pattern recognition. In the activefield of research, numerous classic clustering algorithms havebeen used. However, these algorithms have their owndisadvantages as reported by recent studies. FCM has beenshown to have better performance than HCM. FCM hasbecome the most well-known and powerful method in clusteranalysis. However, these FCM algorithms have considerabletrouble in a noisy environment and inaccuracy with a largenumber of different sample sized clusters. A good clusteringalgorithm should be robust and able to tolerate thesesituations that often happen in real application systems.Here in our work we analyse the data set by using k-meansand fuzzy c-means clustering in which Euclidean Distance isused. After that we use a new metric norm instead ofEuclidean Distance with k-means and fuzzy c-means andanalyse the same data set. After analysis we found that thisnew metric is more robust than Euclidean Norm. These twoalgorithms are called alternative hard c-means (AHCM) andalternative fuzzy c-means (AFCM) clustering algorithms.After analysing these alternative types of c-means clusteringon data set, we found that they have more robustness than cmeansclustering. Numerical results show that AHCM hasbetter performance than HCM and AFCM is better thanFCM as far as time complexity is concerned.
关键词：Hard c-means (or k-means); fuzzy c-means;(FCM); New Metric Norm; Alternative c-means (AHCM;AFCM); Time complexity.