A three-way c-means algorithm

Rough k-means (RKM) is an effective algorithm to deal with uncertain boundary clustering, and it has a variety of extensions since appeared. In RKM and its extensions, data points are coupled with empirical weights to calculate means of clusters, whereas the empirical weights could not necessarily be accurate in all scenarios. In this paper, we propose a three-way weight for each data point according to its inherent characteristics, which can make the weight more accurate. Next, we propose a three-way assignment to assign data points into clusters. By integrating the three-way weight and three-way assignment, we propose a three-way c-means algorithm. Furthermore, in order to validate the effectiveness and efficiency of the proposed algorithm, we compare it with three representative algorithms in the field of RKM, and the proposed algorithm shows the superior performances on certain criteria and statistical measures. (C) 2019 Elsevier B.V. All rights reserved.