A new composite algorithm for clustering problems

This paper presents a composite algorithm for solving a class of clustering problems. A quadratic programming formulation of these problems is considered to be solved by the proposed algorithm. In this algorithm a class of non-convex optimization techniques is applied to an approximated variation of the problem and subsequently a local search scheme is incorporated for the final improvement of obtained solutions. The efficiency of the proposed algorithm is evaluated in comparison with Simulated Annealing and Tabu Search algorithms by exploiting a series of real life data from the literature as well as randomly generated data.