基于案例推理的应急决策贝叶斯网建模方法

贝叶斯网络技术具有丰富的概率表达能力,不确定性问题的处理能力,以及多源信息的融合能力,其有利于辅助突发事件的应急决策,提高决策效率.突发事件应急决策具有较高的时间敏感性,因此要求能够尽可能缩短贝叶斯网络的建模时间,而从无到有的传统贝叶斯网络建模方法的效率显然无法较好地满足这一要求.因此,针对以上问题,提出了基于案例推理的应急决策贝叶斯网络建模方法.该方法基于历史案例库,通过相似度和偏离度两个指标对历史案例进行匹配并得到候选案例,最后通过案例合并和剪枝等方法对候选案例进行调整,最终得到新的案例模型.通过案例仿真对所述方法进行了验证,结果表明:基于案例推理的应急决策贝叶斯网络建模方法没有庞大的搜索空间,也不需要样本数据,只需要提前收集历史案例模型,与传统贝叶斯网络建模方法相比,该方法能够复用历史模型,从而缩短了建模时间,提高了建模效率.

Bayesian network modeling method based on case reasoning for emergency decision-making

Bayesian network has the abilities of probability expression, uncertainty management and multi-information fusion. It can support emergency decision-making, which can improve the efficiency of decision-making. Emergency decision-making is highly time sensitive, which requires shortening the Bayesian Network modeling time as far as possible. Traditional Bayesian net- work modeling methods are clearly unable to meet that requirement. Thus, a Bayesian network modeling method based on case reasoning for emergency decision-making is proposed. The method can obtain optional cases through case matching by the func- tions of similarity degree and deviation degree. Then, new Bayesian network can be built through case adjustment by case merging and pruning. An example is presented to illustrate and test the proposed method. The result shows that the method does not have a huge search space or need sample data. The only requirement is the collection of expert knowledge and historical case models. Compared with traditional methods, the proposed method can reuse historical case models, which can reduce the modeling time and improve the efficiency.