Improving Fuzzy C-Means Clustering by a Novel Feature-Weight Learning

Abstract

Feature-weight assignment can be regarded as a generalization of feature selection. That is, if all values of feature weights are either 1 or 0, feature-weight assignment degenerates to the special case of feature selection. Generally speaking, a number in [0 1] can be assigned to a feature for indicating the importance of the feature. This paper shows that an appropriate assignment of feature-weight can improve the performance of fuzzy c-means clustering. The weight assignment is given by learning according to the gradient descent technique. Experiments on some UCI databases demonstrate the improvement of performance of fuzzy c-means clustering.