作战实验变量因果网络模型检验方法  

A Method for Warfighting Experiment Variables Causal Network Test

作　　者：马心意[1,2] 叶雄兵 刘堃[2] 马雁 MA Xinyi;YE Xiongbing;LIU Kun;MA Yan(AirForce Command College,Beijing 100097,China;Academy of Military Sciences,Beijing 100091,China;Unit 93118,Beijing 100093,China)

机构地区：[1]空军指挥学院,北京100097 [2]军事科学院,北京100091 [3]中国人民解放军93118部队,北京100093

出　　处：《信息工程大学学报》2023年第1期120-128,共9页Journal of Information Engineering University

基　　金：科研基金资助项目(JY2020B012)。

摘　　要：作战实验数据规模大、变量多、因果关系复杂,造成不同变量间因果关系呈现网络化结构特征。在节点数量多、网络结构庞大的情况下,为避免主观经验为检验结果带来偏差,以及盲目开展验证实验产生过度消耗,如何采用量化计算的方式来检验网络化作战实验变量因果关系模型的整体合理性与正确性,是作战实验分析领域亟待解决的重要问题。针对作战实验变量因果网络模型的主要特征,结合因果推断理论方法,按照“调整变量生成—因果效应识别—因果效应异常检验”的基本步骤,提出一种基于因果效应异常检验的作战实验变量因果网络模型检验方法。最终仿真案例表明,该方法能够从已有模型中发现可能存在错误的节点变量,检验结果与问题领域的背景知识较为相符。The large scale of data and complex causality make Warfighting Experiment variables show network characteristics in causal relationship.In the case of a large network scale,in order to avoid bias from subjective experience and excessive consumption from unnecessary validation experiments,it is important to use quantitative calculation to test the rationality and correctness of the Warfighting Experiment variables causal network model overall.Aiming at the main features of Warfighting Experiment variables causal network model,a test method for Warfighting Experiment variables causal network model based on causal effect anomaly test is proposed.It follows three steps of “adjust variable generation-causal effect identification-causal effect anomaly test”.The simulation cases shows that variables can be found in wrong link in Warfighting experiment causal network model.Information we get from simulation cases is agreed with the domain background.

关 键 词：作战实验 数据分析 因果效应评估 因果关系检验 

分 类 号：E919[军事]