基于贝叶斯网络的主效应因素分析

Analysis of Main Effect Factors Based on Bayesian Network

主效应因素分析是故障定位、道路交通事故分析、航空事故分析、风险分析、致因分析等领域关注和研究的重要课题,对于分析事故原因以及采取有效预防措施、降低事故发生率具有重要意义。现有基于贝叶斯网络的致因分析方法多数基于构建的贝叶斯网络结构对影响事故发生的因素进行单独的灵敏度分析,忽略因素之间的交互效应,从而得到片面的结论。本文提出一种基于贝叶斯网络的主效应因素分析方法,在充分利用因素之间的关联关系进行贝叶斯网络结构构建之后,分析各因素对结果因素的影响路径并进行联合影响度分析。实验证明,本文方法克服了简单割裂地对各因素进行单独分析的缺点,能够得到更加可靠全面的致因分析结论。

Main effect factor analysis is an important topic in fault location, road traffic accident analysis, aviation accident analysis, risk analysis, cause analysis and other fields. It is of great significance to analyze the causes of accidents, take effective preventive measures and reduce the incidence of accidents. Most of the existing causal analysis methods based on Bayesian network are based on the constructed Bayesian network structure to conduct a separate sensitivity analysis of the factors that affect the occurrence of the accident, ignoring the interaction effect between the factors, so as to get a one-sided conclusion. In this paper, a main effect factor analysis method based on Bayesian network is proposed. After making full use of the correlation between factors to construct the Bayesian network structure, the influence path of each factor on the result factor is analyzed and the joint influence degree is analyzed. Experiments show that this method overcomes the shortcomings of simple split analysis of each factor, and can get more reliable and comprehensive causal analysis conclusions.