动态时间帕累托最优电子采购多属性谈判模型

Dynamic time-dependent Pareto optimal multi-attribute negotiation model in electronic procurement

基于谈判的电子采购是企业达成供应链交易的重要方式.为了解决采购人员在面对多个谈判属性时决策的困难,本文将多属性谈判问题转化为一个多目标优化模型,采用具有快速非支配排序和精英策略的遗传算法,求解得到帕累托边界,然后设计动态时间依赖策略(DTD),使Agent根据对手出价情况,动态调整让步幅度,在帕累托边界上快速与对手达成交易.另外,对应相关文献中谈判双方互相泄露偏好信息的问题,本文使用基于Agent的自动谈判技术,提出一个有第三方平台参与的自动谈判架构,以及交互协议模型,用以支持自动谈判过程的实施.最后,大量实验证实,相比于领域主流研究成果,本文提出的多属性谈判模型显著提高了谈判效率和成交率,同时也有助于提升谈判双方个体效用和联合效用.

Negotiation-based electronic procurement is essential to enterprises in supply chains. To address decision-making difficulties in a multi-issue negotiation, this study builds a multi-objective optimization model, which is then solved by the genetic algorithm NSGA-Ⅱ for obtaining a Pareto-front. A dynamic timedependent strategy(DTD) is then designed to guide the agent dynamically adjusting its concession along the Pareto-front to make counteroffers according to the opponent’s ever-changing offers. Besides, aiming at the defects of disclosing critical preference information of negotiators in previous studies, this paper uses the agent-based automated negotiation technology to propose an architecture involving a third-party platform and a communication protocol to support the automated negotiation. Finally, the experimental results show that,compared with other studies in the area, our model significantly increases the efficiency and agreement ratio of the negotiation, as well as improving the intrinsic utility and joint utility of both sides.