An improved method using k-means to determine the optimal number of clusters, considering the relations between several variables

Abstract

In this article, we propose a non-hierarchical clustering method that can consider the relations between several variables and determine the optimal number of clusters. By utilizing the Mahalanobis distance instead of the Euclidean distance, which is calculated in k-means, we could consider the relations between several variables and obtain better groupings. Assuming that the data are samples from a mixture normal distribution, we could also calculate Akaike's information criterion (AIC) and the Bayesian information criterion (BIC) to determine the number of clusters. We used simulation and real data examples to confirm the usefulness of the proposed method. This method allows determination of the optimal number of clusters, considering the relations between several variables.