A Hybrid K-Means Algorithm Combining Preprocessing-Wise and Centroid Based-Criteria for High Dimension Datasets

Abstract

Data clustering is an unsupervised classification method aimed at creating groups of objects, or clusters that are distinct. Among the clustering techniques, K-means is the most widely used technique. Two issues are prominent in creating a K-means clustering algorithm—the optimal number of clusters and the center of the clusters. In most cases, the number of clusters is predetermined by the researcher, thus leaving out the challenge where to put the cluster centers so that scattered points can be grouped properly. However, if it is not chosen correctly it will increase the computational complexity especially for high dimensional data set. To obtain an optimum solution for K-means cluster analysis, the data needs to be preprocessed. This is achieved by either data standardization or using principal component analysis on a scale data to reduce the dimensionality of the data. Based on the outcomes of the preprocessing carried out on the data, a hybrid K-means clustering method of center initialization is developed for producing optimum quality clusters which makes the algorithm more efficient. The result showed that K-means with preprocessed data performed better, judging from the sum of square error. Further experiment on the hybrid K-means algorithm was conducted simulated datasets and it was observed that, the sum of the total clustering errors reduced significantly whereas inter distances between clusters are preserved to be as large as possible for better clusters identification.