Journal of Statistical Theory and Applications

ISSN: 1538-7887

LINEX K-Means: Clustering by an Asymmetric Dissimilarity Measure

Authors
Narges Ahmadzadehgoli, Adel Mohammadpour, Mohammad Hassan Behzadi
Keywords
LINEX loss function, dissimilarity measure, k-means clustering
Abstract
Clustering is a well-known approach in data mining, which is used to separate data without being labeled. Some clustering methods are more popular such as the k-means. In all clustering techniques, the cluster centers must be found that help to determine which object is belonged to which cluster by measuring the dissimilarity measure. We choose the dissimilarity measure, according to the construction of the data. When the overestimation and the underestimation are not equally important, an asymmetric dissimilarity measure is appropriate. So, we discuss the asymmetric LINEX loss function as a dissimilarity measure in k-means clustering algorithm instead of the squared Euclidean. We evaluate the algorithm results with some simulated and real datasets.
Download article (PDF)