Abstrakt
K-means clustering is one of the most widely-used partitioning algorithm in cluster analysis due to its simplicity and computational efficiency, but it may not provide ideal clustering results when applying to data with non-spherically shaped clusters. By considering the asymmetrically weighted loss, we propose the K-expectile clustering and search the clusters via a greedy algorithm that minimizes the within cluster tau - variance.
We provide algorithms based on two schemes: the fixed tau clustering, and the adaptive tau clustering. Validated by simulation results, our method has enhanced performance on data with asymmetric shaped clusters or clusters with a complicated structure.
Applications of our method show that the fixed tau clustering can bring some flexibility on segmentation with a decent accuracy, while the adaptive tau clustering may yield better performance. All calculation can be redone via quantlet.com. (C) 2021 Elsevier Inc.
All rights reserved.