Hogeveen H, Noordhuizen-Stassen E N, Tepp D M, Kremer W D, van Vliet J H, Brand A
Department of Herd Health and Reproduction, Utrecht University, The Netherlands.
J Dairy Sci. 1995 Jul;78(7):1430-40. doi: 10.3168/jds.S0022-0302(95)76765-0.
Much specialized knowledge is involved in the diagnosis of a mastitis problem at the herd level. Because of their problem-solving capacities, knowledge-based systems can be very useful to support the diagnosis of mastitis problems in the herd. Conditional causal models with multiple layers are used as a representation scheme for the development of a knowledge-based system for diagnosing mastitis problems. Construction of models requires extensive cooperation between the knowledge engineer and the domain expert. The first layer consists of three overview models: the general overview conditional causal model, the contagious overview conditional causal model, and the environmental overview conditional causal model, giving a causal description of the pathways through which mastitis problems can occur. The conditional causal model for primary udder defense and the conditional causal model for host defense are attached to the overview models at the second layer, and the conditional causal model for deep primary udder defense is attached to the conditional causal model for the primary udder defense at the third layer. Based on quantitative user input, the system determines the qualitative values of the nodes that are used for reasoning. The developed models showed that conditional causal models are a good method for modeling the mechanisms involved in a mastitis problem. The system needs to be extended in order to be useful in practical circumstances.
在畜群层面诊断乳腺炎问题涉及许多专业知识。基于知识的系统因其解决问题的能力，对支持畜群中乳腺炎问题的诊断非常有用。具有多层的条件因果模型被用作开发诊断乳腺炎问题的基于知识的系统的表示方案。模型的构建需要知识工程师和领域专家之间广泛合作。第一层由三个概述模型组成：一般概述条件因果模型、传染性概述条件因果模型和环境概述条件因果模型，对乳腺炎问题可能出现的途径进行因果描述。第二层的初级乳房防御条件因果模型和宿主防御条件因果模型附加到概述模型上，第三层的深层初级乳房防御条件因果模型附加到初级乳房防御条件因果模型上。基于用户的定量输入，系统确定用于推理的节点的定性值。所开发的模型表明，条件因果模型是对乳腺炎问题所涉及机制进行建模的一种好方法。为了在实际情况下有用，该系统需要扩展。
