| Summary: | 碩士 === 淡江大學 === 資訊管理學系碩士班 === 96 === In the combinatorial optimization problem, Traveling Salesman Problem is the most basic and typical example, lots of problems could transfer to TSP as solution. Meanwhile, as the TSP belongs to NP-Complete problem, it is quite hard to find out the best solution in case of huge data, the CTSP, Clustered Traveling Salesman Problem, transferred from a big TSP into several smaller ones, came up. Therefore, it becomes highly important to define a good algorithm in order to find the best similar solution. In consequence, this research proposes an algorithm; the cadre is based on genetic algorithm and tabu search, and transfer the TSP problem into CTSP to find out the best similar solution.
In this research, the algorithm is divided into two parts: the first one on the cluster path, second on the global best path search. The cluster is top- down; at the beginning making the preliminary cluster on the top, then set up the path. The second part, adopting the tabu search: firstly by using the two-points tabu search as large boundary, then by one point tabu search for detail, finally by using MNIO algorithm for the final solution. In TSPLIB international exemplification linkorA100, the path of genetic algorithm is 5.27% longer than the best solution, in this research only 2.46% longer; in the d198, the path of genetic algorithm is 5.96% longer than the best solution, in this research 3.44% only; as for the rat783, the path of genetic algorithm is 10.52% longer than the best solution, here only longer than 8.72%. Therefore, the algorithm of this research can get a better path research result than by using uniquely the genetic algorithm.
|
|---|
