基于SVM的多Agent协商伙伴选择  

SVM-BASED MULTI-AGENT NEGOTIATION PARTNER SELECTION

作　　者：谷琦松[1] 刘胜全[1,2] 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]新疆大学现代教育技术中心,新疆乌鲁木齐830046

出　　处：《计算机应用与软件》2012年第6期131-134,共4页Computer Applications and Software

基　　金：新疆维吾尔自治区科技厅科研项目(200931103)

摘　　要：根据多Agent协商问题的交互特点,引入SVM(Support Vector Machine)分类方法对Agent的协商历史信息进行学习,从Agent的协商历史信息中提取样本来训练SVM,结合模拟协商过程和己方的决策信息,预测与特定伙伴协商时可能出现的结果以及相应的协商收益,根据Agent的自利性原则,选择最合适的协商伙伴。最后,通过仿真实验验证了所提出方法的有效性和优越性。According to the interactive features of Multi-agent negotiation problem,SVM（Support Vector Machine） classification method is involved in to study the Agent’s negotiation history information,extract samples from the Agent’s negotiation history information to train SVM,and combine the simulated negotiation process with one’s decision-making information to predict possible results when negotiating with a particular partner and the corresponding negotiation revenue.Thus,depending on the Agent’s self-interest principle,the most appropriate negotiation partner is selected.Finally,the effectiveness and superiority of the method presented in this paper are verified through simulation experiments.

关 键 词：多AGENT协商 效用 伙伴选择 支持向量机 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]