Study of MDP and K-mediods for TSP problem

Abstract

The classic TSP problem was researched on and CHN144 was chosen to be the data for research. The method that combined MDP and K-medoids was proposed to solve TSP problem in this paper. First of all, cluster CHN144 data through K-medoids and find out the representative objects respectively. Furthermore, the simple TSP problem that consists of representative objects was solved to acquire the optimal path through the Markov Decision Process. Finally, the global optimal path was acquired as 30445km by using the solution above iteratively to the clustering of each object respectively. The feasibility and superiority of this method was proved by analyzing the experiments we conducted in this paper. © 2012 Springer-Verlag Berlin Heidelberg.