Optimal Clustering Algorithms for Data Mining
Omar Y. Alshamesti, Ismail Musa Romi · International Journal of Information Engineering and Electronic Business · 2013
Data mining is the process used to analyze a large quantity of heterogeneous data to extract useful informat ion.Meanwhile, many data min ing techniques are used; clustering classified to be an important technique, used to divide data into several groups called, clusters.Those clusters contain, objects that are homogeneous in one cluster, and different fro m other clusters.As a reason of the dependence of many applications on clustering techniques, while there is no combined method for clustering; this study compares kmean, Fu zzy c-mean, self-organizing map (SOM ), and support vector clustering (SVC); to show how those algorith ms solve clustering problems, and then; compares the new methods of clustering (SVC) with the traditional clustering methods (K-mean, fuzzy c-mean and SOM).The main findings show that SVC is better than the k-mean, fu zzy c-mean and SOM, because; it doesn't depend on either number or shape of clusters, and it dealing with outlier and overlapping.Finally; this paper show that; the enhancement using the gradient decent, and the proximity g raph, imp roves the support vector clustering time by decreasing its computational complexity to O(n logn) instead of O(n 2 d), where; the practical total time fo r improvement support vector clustering (iSVC) labeling method is better than the other methods that improve SVC.