A Game Theory Framework for Clustering

Abstract

Abstract. The Game Theory-based Multi-Agent System (GTMAS) of Toreyen and Salhi, [10] and [12], implements a loosely coupled hybrid algorithm that may involve any number of algorithms suitable, o priori, for the solution of a given optimisation problem. The system allows the available algorithms to co-operate toward the solution of the problem in hand as well as compete for the computing facilities they require to run. This co-operative/competitive aspect is captured through the implementation of the Prisoners' Dilemma paradigm of game theory. Here, we apply GTMAS to the problem of clustering European Union (EU) economies, including Turkey, to find out whether the latter, based on a number of criteria, can fit in the EU Eind find out which countries, if any, it has strong similaries with. This clustering problem is first converted into an optimisation problem, namely the Travelling Salesman Problem (TSP) before being solved with GTMAS involving two players (agents) each implementing a standard compbinatorial optimisation algorithm. Computational results are included.