A new weighting fuzzy c-means algorithm

Jian Yu, Houkuan Huang

Abstract: It is always supposed that each cluster has almost equal number of points when carrying out the FCM. However, this assumption may not hold in practice. In this paper, we propose a novel fuzzy c-means algorithm based on the definition of generalized mean, weighting fuzzy c-means algorithms (WFCM). Noticing that the Gath-Geva fuzzy clustering algorithm considers the size of the clusters, we compare the performance between the WFCM and the Gath-Geva fuzzy clustering algorithms by numerical experiments. Moreover, we offer a theoretical threshold to choose an appropriate weighting exponent in the WFCM, which is also valid for the FCM, the numeric experiments verify such conclusion.