【作者】 周云；

【导师】 傅泽田；

【作者基本信息】 中国农业大学 ， 管理科学与工程， 2005， 硕士

【摘要】 本文针对传统鱼病诊断专家系统知识获取的瓶颈问题利对诊断思维模拟存在的不足,提出用基于案例推理的研究方法来研究鱼病诊断问题。一方面,基于案例推理的研究方法不需要对专家知识进行规则化处理,重要的是获取足够的鱼病案例;另一方面,基于案例的推理是模拟鱼病诊断中形象思维的有效工具。 全文围绕鱼病案例的分析、表示、存储与检索展开讨论。首先,在全面分析鱼病诊断领域知识的基础上,探讨了鱼病诊断中的思维模式,并综合鱼病诊断专家系统各案例元素间的关系,给出了基于案例推理的鱼病诊断概念模型。进而依照关系数据库的基本理论设计了鱼病诊断问题案例库,确定了合适的案例表示方式以及存储结构。 在讨论案例检索策略时,针对鱼病案例属性定性、离散化的特点,将鱼病案例转化为用二值逻辑表示,进而探讨了基于相似性度量模型利基于粗分析方法的两种案例检索策略。前者在改进TC相似性度量模型的基础上实现了一种动态的变权机制;后者通过相对约简建立多级索引来完成案例的检索。两种检索策略从不同角度处理了鱼病诊断问题中存在的不确定性。 最后,根据开发专家系统的技术要求,在Windows2003 Server+Apache+Tomcat5.0系统/客户开发平台下,采用JSP+JavaBeans的应用层模式,建立了基于WEB的B/S/S三层体系结构的鱼病诊断专家系统。更多还原

【Abstract】 In this paper, Case-Based Reasoning is put forward to study fish disease diagnosis in order to solve the problem of knowledge acquisition bottleneck existing in the previous fish disease diagnosis expert systems. The advantage of CBR is following: 1) it just needs adequate cases and the regularization and pretreatment of expert knowledge is not necessary. 2) CBR is an efficacious method to simulate expert thinking with imagery.The paper discusses the presentation of a new case, the retrieval of the most similar cases from the case base, and the storage of all cases. Above all, we analyze the diagnosis process offish disease the fish expert have, and put forward a conception model of fish disease diagnosis according to the relation of case elements. Furthermore, we design the case representational formalism and the structure of case base according to the relation database theory.Fish cases are qualitative and discrete data, in view of these characteristics, binary system is used to represent the cases when we discuss the case retrieval strategy. On the basis of the above, two case retrieval methods are provided, one is similarity-testing model, and the other is rough analysis. The similarity-testing model realizes a mechanism of variant weights of the cases according to Tversky’s contrast matching model. In the second method, rough analysis is introduced, which can fully use the reduction attributes of case database and make multi-index of case. The two case retrieval methods resolve the uncertainty offish disease from different points of view.Finally, according to the demand of expert system, the model of JSP added JavaBeans has been developed for implementing the fish disease diagnosis expert system under Windows 2003 Server. A three layers structure B/S/S used in WEB is adopted in the new fish disease diagnosis expert system.更多还原