Abstract:

An improved fuzzy C-means (FCM) clustering algorithm based on intuitionistic fuzzy sets is proposed for a set of multi-feature clustering analysis problems, in which feature weights are unknown and feature values are intuitionistic fuzzy numbers. A regional density parameter is defined, and c samples with the farthest Euclidean distance from the high-density region are selected as initial clustering centers. The Characteristic values are weighted with Characteristic weights calculated by using intuitionistic fuzzy entropy. Calculation steps for the improved FCM clustering algorithm based on intuitionistic fuzzy sets are given. Validity of the proposed improved algorithm is checked with a numerical example. The improved algorithm can solve the problem of falling into local minima, and greatly reduce iterative time so as to accelerate convergence.

Key words: fuzzy C-means clustering algorithm, intuitionistic fuzzy set, fuzzy entropy, region density, initial clustering center