民航安全政策主题挖掘与成效评估分析  

Topic Mining and Effectiveness Evaluation Analysis of Civil Aviation Safety Policy

作　　者：程明 杨迪木 CHENG Ming;YANG Dimu(School of Safety Science and Engineering,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学安全科学与工程学院,天津300300

出　　处：《安全》2025年第3期88-96,共9页Safety & Security

基　　金：国家重点研发计划(2023YFB4302901);民航安全能力项目(ASSA2023/19)。

摘　　要：为分析现阶段民航安全政策体系的治理效能,基于2016—2023年民航局发布的86份民航安全类政策文件,利用隐含狄利克雷分布(LDA)主题模型挖掘政策文本主题,从共现网络视角探究政策文本主题及关键词之间的关系;采用斯皮尔曼相关系数分析法,系统分析民航安全政策与运输航空征候原因之间的相关性。结果表明:当主题困惑度值为4时,民航安全类政策划分为应急管理、地面运行、风险管理和飞行运行4个政策文本主题,涵盖民航安全运行重点关注的领域;共现网络显示安全政策以跑道和安全隐患为双核心网络;结合2016—2023年运输航空征候数据评估安全政策的效能,发现4类政策在应对因天气意外、空管和机械原因所致征候时效果明显,但在因机务、地面保障、机组、航务和管理原因所致征候时作用相对较小。可见,民航安全类政策文本主题挖掘模型可为民航安全政策制定和态势分析提供量化支持。In order to analyze the governance effectiveness of the current civil aviation safety policy system,based on 86 civil aviation safety policy documents issued by the Civil Aviation Administration from 2016 to 2023,the policy text topics were mined using the Latent Dirichlet Allocation(LDA)topic model,and the relationship between the policy text and the keyword was explored from the perspective of the co-occurrence network.Spearman correlation coefficient analysis was applied to systematically analyze the correlation between safety policies and causes of transport aviation incidents.The results show that civil aviation safety policies were classified into four policy text topics:emergency management,ground operations,risk management,and flight operations when the topic perplexity value is 4,encompassing the focused domains of the aviation operation safety;that the co-occurrence network indicates that the safety police takes runways and safety hazards as the dual core network;and that the four policy types present significant effectiveness in addressing incidents caused by unexpected weather,air traffic control,and mechanical factors,but the effect is little relatively in addressing incidents caused by maintenance,ground support,crew operations,flight dispatch,and administrative issues,based on the effectiveness of applying transport aviation incident data to access safety policies from 2016 to 2023.The thematic mining model of civil aviation safety policy texts can provide quantitative support for both policy formulation and situational analysis in the aviation safety management.

关 键 词：主题挖掘 安全类政策 隐含狄利克雷分布(LDA)主题模型 共现网络 

分 类 号：X949[环境科学与工程—安全科学]