A new semi-supervised clustering algorithm with pairwise constraints by competitive agglomeration

Recently semi-supervised fuzzy clustering with pairwise constraints was developed, in which the disagreement on the magnitude order between penalty cost function and the basic objective function will cause over adjustment of membership values and their deviation from the normal range. In order to solve this problem, an improved semi-supervised fuzzy clustering algorithm with pairwise constraints (SCAPC) was proposed based on a redefined objective function. The new penalty cost function in SCAPC theoretically conforms to the methodology of classical fuzzy clustering, which is expressed as the violation cost incurred by the pairs, and has the same magnitude order as the basic objective function. Experimental results on benchmark datasets and images showed that SCAPC can produce more accurate clustering by moderately enhancing or reducing the ambiguous memberships. Research indicates that constraint term of the proposed algorithm can achieve a good agreement and cooperation with the basic objective function. (C) 2011 Elsevier B.V. All rights reserved.