Double‐weighted fuzzy clustering with samples and generalized entropy features

Abstract

For the dataset with different sample contributions and different feature importance, it is difficult to acquire a proper cluster structure that covers the entire features of the sample set. To improve the clustering result, a novel weighted fuzzy clustering algorithm based on both samples and generalized entropy features, called SGEF‐WFCM, is proposed in this article, among which a new objective function is developed on the basis of feature‐weighted generalized entropy regularization with a double‐weighting strategy of samples and features, the weighted coefficients of the features to each cluster, as well as the importance of the samples to the cluster are calculated dynamically, to obtain a better clustering result. Finally, experiments on both synthetic datasets and real‐world datasets from UCI are employed to verify the performance of the proposed SGEF‐WFCM algorithm. The results show that SGEF‐WFCM is superior to the conventional FCM algorithm in both the effectiveness and the usefulness during practices.