基于贝叶斯网的专家系统平台的开发与应用

Development and application of expert system platform based on Bayesian network

针对基于贝叶斯网建立专家系统所涉及的关键问题,采用分离技术、"噪音或"技术和转换现有的确定性因子知识库等技术简化贝叶斯网的构造,利用变量消去算法实现了贝叶斯网的精确推理,开发了贝叶斯网建网组件和推理组件,并应用这两个组件建立了简单的奶牛疾病诊断专家系统和蔬菜种植专家系统。实际运行表明,所开发的贝叶斯网建网组件和推理组件在表示与处理不确定性知识时是非常有效的。

The methods and technology of applying Bayesian network to build expert system are introduced, especially the construction and inference of Bayesian network. In the construction of Bayesian network, noisy - or technology, divorcing method and transformation from certainty factor model to Bayesian network are used. As for the reasoning method, variable elimination algorithm is used. Based on the above method and technology, the Bayesian construction component and reasoning component are developed, which then are used to build cow disease diagnosis system and vegetable planting system. In running the two systems, the result is conformed with the domain expert, which proves that it is effective to use the Bayesian network constructing component and reasoning component to represent and deal with uncertain knowledge.