基于半定性贝叶斯网络的建模方法研究

在定性贝叶斯网络的基础上,提出一种既包含定性符号信息,又包含概率分布定量信息的半定性贝叶斯网络,用以集成多个专家的知识和经验,通过一个与专家反复交互的过程,构建出相关问题领域的稳定的贝叶斯网络模型。提出利用区间数来统一表达多个专家的定性或定量的判断意见,给出了在半定性贝叶斯网络中进行概率推理的区间数传播算法,并通过算例验证其有效性。

Study on the Modeling Method Based on Semi-qualitative Networks

On the basis of qualitative Bayesian networks, a semi-qualitative networks is proposed, which include both qualitative sign and quantitative probability distribution. The semi-qualitative networks are used to aggregate multi-experts' knowledge and experience by a stepwise procedure, so as to construct the Bayesian networks in the domain. interval to uniformly express qualitative and quantitative knowledge of multi-experts is used. An interval propagation algorithm to perform probability inference is also present in the semi-qualitative networks. Finally, an example to validate is given.