一种基于Metropolis准则遗传算法的多边多议题协商  

A Genetic Algorithm Based on Metropolis Rule and Its Application in Multi-Lateral Multi-Issue Negotiation

作　　者：李剑[1] 景博 杨义先[1] 

机构地区：[1]北京邮电大学网络与交换技术国家重点实验室信息安全中心,北京100876 [2]北京应用气象研究所计算机室,北京100029

出　　处：《电子学报》2009年第5期1037-1040,共4页Acta Electronica Sinica

基　　金：国家973项目(No.2007CB310704);国家自然科学基金(No.90718001)

摘　　要：为了使得基于智能体电子商务中协商智能体最大限度地达到协商的满意解,甚至最优解,并且提高协商的效率,提出将基于Metropolis准则的遗传算法应用于基于智能体电子商务的多边多议题同时出价的协商当中.经过1000次的实验数据表明,简单遗传算法平均需要155次才能达到协商的最优解,而基于Metropolis准则的遗传算法平均需要76次就达到协商的最优解.这个结果表明,基于Metropolis准则的遗传算法可以使得多边多议题协商中的智能体高效达到协商的最优解.To make the negotiation agents gain satisfying result and negotiate more efficiently in multi-agent based e-commerce, a genetic algorithm based on metropolis rule is presented and the algorithm is applied in multi-lateral multi-issue simultaneous bidding negotiation.After 1000 times of experiments,simple genetic algorithm averagely needs 155 runs to gain the satisfying result, while the genetic algorithm based on metropolis rule averagely needs 76 runs to gain the satisfying result. The experimental result show the genetic algorithm based on metropolis nile can gain the optimal negotiate result more efficiently in multi-literal multi-issue negotiation.

关 键 词：电子商务 多边多议题协商 遗传算法 METROPOLIS准则 

分 类 号：TN301[电子电信—物理电子学]