The LINEX Weighted k-Means Clustering

Abstract

LINEX weighted k-means is a version of weighted k-means clustering, which computes the weights of features in each cluster automatically. Determining which entity is belonged to which cluster depends on the cluster centers. In this study, the asymmetric LINEX loss function is used to compute the dissimilarity in the weighted k-means clustering. So, the cluster centroids are obtained by minimizing a LINEX based cost function. This loss function is used as a dissimilarity measure in clustering when one wants to overestimate or underestimate the cluster centroids which helps to reduce some errors of misclassifying entities. Therefore, we discuss the LINEX weighted k-means algorithm. We examine the accuracy of the algorithm with some synthetic and real datasets.

Keywords

• LINEX loss function
• Feature weights
• Weighted k-means
• Clustering