面向Agent马尔科夫多议题自动谈判模型研究  

The Model of Agent-oriented Multi-issue Automated Negotiation Based on Markov Chain Method

作　　者：陈培友[1,2] 高太光[1] 李一军[2] 

机构地区：[1]黑龙江科技学院,黑龙江哈尔滨150027 [2]哈尔滨工业大学,黑龙江哈尔滨150001

出　　处：《管理工程学报》2014年第3期203-208,共6页Journal of Industrial Engineering and Engineering Management

基　　金：黑龙江省博士后基金资助项目(LBH-Z05129);国家自然科学基金资助项目(71071028);黑龙江省自然科学基金资助项目(G201021);黑龙江省骨干教师创新资助项目(1154G19)

摘　　要：为增强自动谈判模型中各谈判Agent在报价与反报价过程中的科学性和合理性,提高谈判的效率和成功的概率,将马尔科夫链方法应用于对谈判对方各轮报价的预测当中,以增强模型对谈判过程的逻辑推理能力,通过与改进后的遗传算法相结合,构建了面向Agent的马尔科夫多议题自动谈判模型,期望为谈判各方在谈判过程中的报价提供有力支持。通过仿真计算可以看出,所提模型很好地对谈判过程进行了仿真和推理,多种方法的有效结合,能够客观地为谈判各方Agent提供合理决策的有效依据。A growing number of conflicts are seriously restricting the healthy and orderly development of our society and economy.Negotiation is an effective means to solve differences and disputes among people in a society.The online negotiation is indispensable to the success of electronic commerce.However,modeling and evaluating negotiation is very challenging because many problems in negotiations are uncertain and many important factors need to be considered.According to the characteristics and requirements of automatic negotiations,this paper applies the Markov Chain Method （MCD）to speculate the offers of buyer in order to improve scientific and rationality of each offer in an auto-negotiation model.In addition,MCD method can enhance the efficiency and success of negotiations because the method can reveal the future development trend using state transition probability.Moreover,this method can reflect the degree of influence of all random factors and inner connections between different states based on the change of transition probability.Negotiations conditions are constantly adjusted along with the negotiations process.The changes of attribute values in a concrete negotiation process are only related to recent negotiation states rather than the past states.Therefore,negotiations can be approximately modelled as a Markov Chain （MC）.Moreover,MC is applied to predict the change of negotiation attribute values.Negotiation data mainly come from each side in negotiations and the negotiation predictions are based on the previous negotiation results.The calculation process varies according to states transfer cycle changes.Thus,each negotiation cycle can have certain credibility as long as the initial transfer matrix of data is accurate.In order to select the best negotiation decision,this paper combines MCD and Improved Genetic Algorithm （IGA） to achieve global optimization and provide a reference for all parties involved in making scientific decision.Our automatic negotiation model selects the negotiation tactic t

关 键 词：多议题自动谈判 马尔科夫链方法 面向AGENT 改进遗传算法 信息熵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]