事例ベース推論に基づく巡回セールスマン問題の一解法

抄録

It is difficult to obtain the optimal solution to combinatorial problems. The more the number of variables increases, the more calculation time explosively increase. Therefore approximation methods are needed to the kind of problems. Traveling Salesman Problem (TSP) is one of typical combinatorial problems. Using geometric information, this problem can be solved effectively. It is difficult however to use the information on the basis of almost all approximation method. It is shown that the information can be used effectively in Case -Based Reasoning (CBR) system to TSP.
In this paper, we make a system based on CBR for TSP and present a new retrieval method to find the most similar case for the problem. In addition, by dividing a large combinational problem into sub-problems and applying the system to each sub-problem, an approximation solution to a large combinational problem can be obtained in ten or so minutes.