基于Agent的情感劝说的信任识别模型研究

Research on trust recognition model of emotional persuasion based on Agent

由于Agent系统的分布式特征,使得传统的安全手段已经不适用于新的应用需求,而解决该问题最重要的思路是信任管理,其基础之一就是信任识别。本文在Agent情感劝说的背景下,利用多维度评价信息和Agent关系网络,运用Dempter-Shafer证据理论,提出基于Agent的情感劝说的信任识别模型,引入动态权重因子,将直接交互信息和推荐信息进行组合,计算出各合作伙伴的综合信任度值,从中寻找出适合的合作伙伴,为Agent的决策提供大力的支持。

With the development of Internet technology and economy, transactions between organizations have become more and more frequent. As a necessary business activity, negotiation has become more and more important. Information, as an important resource, plays a vital role in the success of the negotiation.The negotiation mode based on Agent has greatly improved the efficiency of negotiation because of its integration into advanced information technology.However, because of the openness and uncertainty of the Agent system, the Agent system has also encountered many problems at the same time, and trust is one of the most important problems. In multi-Agent systems, Agent often represents different organizations or users, and in order to succeed, these Agent need to cooperate, coordinate and negotiate with other Agent to achieve their goals. This requires that the multi-Agent system can measure the trust degree of the interactive side, and the trust value of Agent is calculated by the trust model. The trust recognition model is established in this paper, and the appropriate partner is selected by this model.Under the background of Agent’s emotional persuasion, using the multi dimension evaluation information and the Agent relation network and using the Dempter-Shafer evidence theory, the trust recognition model of emotional persuasion based on Agent is put forward, and the dynamic weighting factor is introduced to combine the direct interaction information and the recommendation information to calculate the comprehensive trust value of the partners. From that, we can find out suitable partners to provide strong support for Agent’s decision making. The validity of the model is proved by an example.With the direct interaction history between Agent and the recommendation information of other Agents on the target Agent in the Agent relationship network, direct trust and indirect trust are combined with dynamic weighting factors. In computing trust, trust is considered as a combination concept, and the persuasion and emotional factors of target Agent in the interaction of the past are introduced. In comparison with human studies, the improvements and conclusions drawn from this study are summarized as follows:(1) Trust is a combinatorial concept and uses multi-dimensional evaluation information. This article is based on the ontology of trust, that is, trust is composed of many factors. The main factors that affect the trust of both parties are the quality of supply, the degree of supply completion in the prescribed time,the persuasion in the transaction and the pleasure of the cooperation between the two parties in the transaction, and the four dimensions of the transaction are fully utilized. The evaluation information, which completes the trustworthiness measurement of the service providers, improves its applicability and flexibility.(2) Direct trust and indirect trust are combined by dynamic weight factors, which is more close to reality. Direct trust and indirect trust are considered in the collection of trust information, and the dynamic weighting factor D is introduced in the combination of the two, that is, the weight is different with the number of direct interaction between the two parties. For example, when the buyer and target Agent have sufficient interaction experience, they will tend to give direct trust a larger weight. The combination of dynamic weight factors can make the results closer to reality.(3) Introducing emotion and persuasion influence factors to make the model more intelligent and adaptable to dynamic and complex environment. In real life, persuasion must be used in the process of negotiation. The strength of persuasion is also a strong degree of persuasion. The pleasure of the two parties also shows the satisfaction of the buyer with the results of the cooperation. The two factors are introduced into the trust model to make the model consider the emotional factors in the calculation. The calculation results are more practical.(4) Using Agent relational network to find recommender, to a large extent, avoid malicious recommendation. In this paper, the Agent relationship network is used to find the referer, because the information mobility of the Agent network is large, the information receiver is the potential collaborator of the two parties,and the breach of the cooperative party or the behavior that does not conform to the standard will affect the view of the transaction party in the next cooperation.Therefore, in the Agent relations network the default cost of Agent is very large, to a certain extent, and it can avoid malicious recommendation or breach of contract in cooperation.(5) It is a simple and effective way to get the trust evaluation by first hand material after grading each attribute of every transaction. After each transaction,key attributes, such as the quality of the service or goods, the degree of delivery completion within the specified time, the persuasion in the process of transaction, the degree of pleasure in the transaction process and the outcome, and the hierarchical information accumulated over a number of services to obtain the trust evaluation of the target Agent, can be used to improve partner selection efficiency.