【作者】 符一行；

【导师】 姚莉；

【作者基本信息】 国防科学技术大学 ， 管理科学与工程， 2012， 硕士

【摘要】 随着网络技术的出现和信息集成技术的快速发展,辩论技术在决策领域的应用发挥了重要的角色。尤其是当前决策者面临信息过载,以及集成信息具有有动态性、不完整性、不一致性等特点,迫切需要辩论作为一种新的技术手段辅助决策。一方面因为辩论通过展现论据之间一种支持或攻击关系,可以为决策提供理由和支撑;另一方面辩论能够通过论据的构建和评估对已选方案进行解释和分析说明。从而使决策者深入了解动态环境中决策所面临的多种因素和不确定性,帮助决策者正确判断可能采取的行为,从而有效改善决策的效果。本文首先研究Dung的抽象辩论框架和基于价值的辩论理论,并总结分析当前基于价值辩论决策支持模型存在的不足与缺陷,提出了论据分类的思想,然后根据该思想设计了ArguDecision辩论决策控制算法,通过治疗血液凝稠实例对算法进行了验证,充分证明了该算法的可行性与有效性。但是从目前调研的文献可以看出使用论据制定和解释决策,这类决策模型需要先依据Dung框架求出每一方案的所有可接受论据,然后,再依据不同类型的论据比较规则说明决策方案之间的优劣。由于基于Dung框架计算可接受论据是NP-Completeness问题,该方法的计算实现依然是一个难以解决的困难问题。因此本文针对动态不确定组织环境,提出了一种基于辩证思维的决策支持方法,我们将该方法称之为用于辅助决策的辩证分析模型DAM-DS(Dialectic Analyzing Models for Decision Support)。该方法将决策过程分为两个阶段:决策方案的可行性论证阶段和决策方案的选优阶段。并且结合辩证法和BDI模型的优越性,该方法将Katie等人提出的论据模式和表示Agent心智状态的BDI Agent计算模型相结合建立论据的结构,同时,扩展了传统的基于对话博弈的辩论框架,明确表示了辩论参与者的不同信念和论据的辩证语义,基于辩证语义(Dialectic Semantics)实现决策方案的可行性评估和决策方案的选优,并能通过论据解释决策的理由。本文基于DAM-DS模型,设计实现了Smart doctor医疗方案辅助决策系统,并将该系统应用到血液净化治疗方案的辅助决策中,通过该系统证明了该模型在辅助决策中是可行的。更多还原

【Abstract】 With the emergence of IT and rapid development of Information Integration technique, argumentation technology has played important role in Decision-Support Domain. Especially for information overload and dynamic, incomplete, inconsistent information environment, it is urging needed to adopt argumentation technology to support decision. On the one hand, argumentation may provide reasons and supports for decision through showing support or attack relation between arguments; On the other hand, argumentation can explain and analyse chosen options through constructing and evaluating arguments. Since this, it helps decisionmaker to estimate and judge actions, and it will improve decision effect.Firstly this dissertation makes research on Dung abstract argumentation scheme and argumentation theory based on value, it summarises the flaw and shortcomings of current Value-based argumentation decision-support model. It puts forwad the theory of argument category, In the light of it we designed the Argumentation-Decision controlling algorithm ArguDecision. We have implemented it to validate the feasibility and effectivity of this algorithm. However this kind of decision models needs to compute preferred extension repeatly. Since computing preferred extension is NP-Completeness problem,it is relatively difficult to solve this problem.This dissertation proposes a Dialectic Analyzing Models for Decision Support, it seperates the decision process into two phases: evaluating defensible options phase and picking best option phase. It integrates Katie’s Argument scheme and BDI Agent model. We may evaluate options and choose better option on the basis of Dialectic Semantics.This paper has designed Smart doctor Medicalcare decision support system, and implemented it in the Blood Purification Domain.Finally we has proved that this model is available to be applied in decision support.更多还原