基于BBN的航空公司风险评估技术研究

Research on Risk Evaluation for Airlines Based on BBN

在人、机、环境和管理(MMEM)系统理论的基础上,通过研究有关民航规则、标准中的相关内容建立航空公司风险评估指标体系;利用层次分析法(AHP)确定指标权重,运用模糊综合评价法(FCE)量化其中的定性指标;基于贝叶斯置信网络(BBN)软件GeNIe完成了整个系统贝叶斯网络拓扑结构的建立,将航空公司历史数据作为贝叶斯网络参数学习的训练数据,获取节点概率表(NPT),建立航空公司风险评估模型,并对其进行概率推理计算;通过环比分析确定影响风险的主要因素。

On the basis of the theory of MMEM （Man, Machine, Environment and Management） system, an airlines risk evaluation index system is established through a research on civil aviation rules and standards. The weights of indexes are determined by using the analytic hierarchy process （AHP）, and the qualitative indexes are quantified by using fuzzy comprehensive evaluation （FCE）. Based on a Bayesian belief network （BBN） software-GeNie, the whole system of the Bayesian network topology structure is established, and by using the historical data of airlines as Bayesian network parameter learning training data, a node probability table （NPT） is obtained. Thus an airline risk evaluation model is built to carry on the inference of probability calculation and determine the main impact factors on safety through chain analysis.