CLUSTERING AND CLIQUE PARTITIONING - SIMULATED ANNEALING AND TABU SEARCH APPROACHES

We study the application of simulated annealing and tabu search to the solution of the clique partitioning problem. We illustrate the effectiveness of these techniques by computational results associated not only with randomly generated problems, but also with real-life problems arising from applications concerning the optimal aggregation of binary relations into an equivalence relation. The need for these approaches is emphasized by the example of a special class of instances of the clique partitioning problem for which the most commonly used heuristics perform arbitrarily badly, while tabu search systematically obtains the optimal solution.