基于Multi-Agent的多属性拍卖评标机制的仿真研究

Simulation Research of Bidding Evaluation Mechanism in Multi-Attribute Auction Based on Multi-Agent

针对当前分组多属性评标机制,在评标专家有限理性的前提下,存在着评标专家抬分和压分越来越严重、对立情绪累积现象明显和组间评标结果差异越来越大的问题。提出加入综合专家组或增设一对反映标的负相关新指标的策略改进当前评标机制,描述了改进评标机制的原型系统,构建了基于Multi-Agent的分组评标行为仿真构架,在定义各Agent的属性、决策规则和交互协作方式基础上,给出了仿真的具体步骤,定义了相关行为参数,利用Matlab GUI平台实现评标行为仿真分析系统,设定相关参数进行三种机制下的评标模拟与比较,仿真结果表明:改进后的分组评标机制有效地控制了评标结果差异增大,相对合理地调节了评标专家评分标准,提高了评标公平性,体现了改进机制的有效性,为评标决策提供依据。

In the practices, the bid attributes usually are classified into two groups, technical attributes and business attributes. Taking bounded rationality as precondition, there would be that raising and pressing grades was increasingly severe, the difference between two groups increasingly and objectiveness worse and worse. So a group of integrative experts were set or some criteria related with two groups of attributes were added to improve current mechanism. Prototype system was described, meanwhile the multi-agent simulation frameworks were structured. After the attributes of each agent, decision rule and the way were defined in which different agents used for exchange and collaboration, the simulation procedure was given specifically. Then the analytical system of bidding evaluation based on Matlab GUI was realized after defining correlation behaviour parameters. The comparison results demonstrate that two improved mechanisms could adjust standard for evaluation reasonably, could control the difference increasing between two groups effectively and improve the fairness of evaluation. It indicates the effectiveness of improved mechanisms.