摘要
基于历史管制指令偏差案例的分析,案例风险特征的挖掘方法以提高案例匹配的准确性。建立概念向量模型,解决文本词义相近产生冗余特征的现象,构建风险模式案例库。考虑传统欧式距离无法区分属性差异化,提出基于粗糙集理论的加权欧式距离进行案例相似度匹配,以高权重表征突出人为因素偏差、风险指令模式关键风险特征的重要性,来提高案例匹配的准确性。对比试验结果显示,基于粗糙集理论的加权欧氏距离能准确区分出关键风险特征,且案例间的距离由原来的0.2~0.4缩小到0.1~0.2,提高了匹配的准确度。并进行实时案例仿真,结果验证了改进后的方法适用于属性差异化大的案例匹配。
Analyze the cases of deviations of historical control instructions,and study how to further explore the risk characteristics of cases and improve the accuracy of case matching.Based on the text mining technology,the features of the risk cases are extracted,and the concept vector model is established to solve the phenomenon of redundant features that are similar in the meaning of the text.The attribute values of the risk features are converted from text to numeric to construct a risk model case library.Consider that the traditional Euclidean distance cannot distinguish attribute differences,a weighted Euclidean distance based on rough set theory is proposed for case similarity matching.The priority matching of key risk features with high weights such as human factor deviation and risk instruction mode is adopted to improve the accuracy of case matching.The comparison test results show that the weighted Euclidean distance based on rough set theory can accurately distinguish the key risk features,and the distance between cases is reduced from the original 0.2—0.4 to 0.1—0.2,which improves the accuracy of matching.Through real-time case simula⁃tion,the results verify that the improved method is suitable for case matching with large attribute differences.
作者
毛继志
叶海生
吴磊
汤新民
郭鸿滨
MAO Jizhi;YE Haisheng;WU Lei;TANG Xinmin;GUO Hongbin(China Institute of Aeronautical Radio and Electronics,Shanghai 200241,China;Civil Aviation College,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China;Air Traffic Management Bureau of Civil Aviation Administration of China,Beijing 100022,China)
出处
《交通信息与安全》
CSCD
北大核心
2020年第4期50-57,共8页
Journal of Transport Information and Safety
基金
国家自然科学基金项目(61773202)
四川省科技计划项目(2018JZ0030)
中国航空无线电电子研究所航空电子系统综合技术国防科技重点实验室基金项目(6142505180407)资助。
关键词
管制指令偏差事件
案例匹配
粗糙集理论
加权欧式距离
regulatory directive deviation event
case matching
rough set theory
weighted euclidean distance
