基于多智能体系统的分布式信誉机制研究

发布时间：2018-03-17 15:23

本文选题：多智能体系统　切入点：信任　出处：《管理工程学报》2013年01期 　论文类型：期刊论文

【摘要】：基于多智能体系统原理,提出了分布式信誉机制中信誉信息的获取与集结模型。在分布式信誉系统中,信任方可以通过三种途径获取直接或间接的信誉信息:与被信任方的直接交易经验,借助自身社会网络找到与被信任方有过交易的见证者,以及要求被信任方提供与其有过交易的引荐者。首先,引入了一个激励机制,促使被信任方诚实地提供引荐者信息,保证信誉信息获取的数量和质量。其次,考虑到不同来源的信誉信息具有不同的可信度,在信誉信息集结时,基于用户相似度计算可信见证者的评价可信度,基于陌生引荐者和见证者的评价与可信见证者的评价集结结果之间的结果相似度计算陌生引荐者和见证者的评价可信度,建立了信誉集结模型来计算被信任方的综合信誉度。通过实验仿真,表明该模型具有较强的有效性和健壮性,其不但可以使信任方获取足够多的信誉信息,而且可以有效地削弱恶意见证者和引荐者诋毁尤其是共谋的影响。
[Abstract]:Based on the principle of multi-agent system, this paper proposes a model of obtaining and aggregating reputation information in distributed reputation mechanism. Trust parties can obtain direct or indirect credit information through three ways: direct trading experience with trusted parties, and use of their own social networks to find witnesses to transactions with trusted parties. First, an incentive mechanism is introduced to enable the trusted party to provide the referrer information honestly and to ensure the quantity and quality of the credit information obtained. Considering that credibility information from different sources has different credibility, the credibility of trusted witnesses is calculated based on user similarity when credit information is assembled. Based on the similarity between the evaluation of strange introducers and witnesses and the aggregate results of credible witnesses, the credibility of the evaluation of strangers and witnesses was calculated. The credibility aggregation model is established to calculate the credit degree of the trusted party. The simulation results show that the model is effective and robust, and it can not only make the trust party obtain enough credit information, Moreover, it can effectively weaken the influence of malicious witnesses and referrals, especially complicity.
【作者单位】：西安交通大学管理学院;
【基金】：国家自然科学基金重大资助项目(70890081)
【分类号】：TP393.08;TP18

【参考文献】