基于文本挖掘的管制运行风险主题分析*

为充分挖掘管制运行风险信息和隐藏规律，实现数据驱动的风险管理。以某管制单位2004—2019年共269条管制原因不安全事件数据为挖掘语料，在考虑上下文语义的基础上，运用潜在狄利克雷分配（LDA）主题模型挖掘管制运行风险主题及关键词，使用Word2Vec挖掘主题之间、关键词之间的关联关系，运用社会网络分析软件UCINET、可视化工具NETDRAW构建语义网络将关联关系进行可视化并进行网络分析。结果表明:LDA主题模型可以通过运行数据实现对管制运行风险的高效提取和深层挖掘，挖掘到管制人为因素、特情处置、地空配合、班组资源管理、组织管理、运行环境、管制指挥共7个主题，其中管制人为因素主题是核心主题，与其他主题都具有较强的相互关联；Word2Vec和语义网络相结合能够更准确地挖掘风险之间的关系，确定主题的重要度排序，识别关键风险。

Analysis on operation risk topics of air traffic control (ATC) based on text mining

In order to fully mine the operation risk information and potential rules of air traffic control (ATC),and realize the data driven risk management,taking the data of 269 ATC caused unsafe events in a ATC unit from 2004 to 2019 as the mining language materials,on the basis of considering the context semantics,the Latent Dirichlet Allocation (LDA) topic model was used to mine the topics and keywords of ATC operation risks.The Word2Vec was applied to mine the association relationships between the topics and keywords,and the social network analysis software UCINET and the visualization tool NETDRAW were used to construct a semantic network,visualize the association relationships and perform the network analysis.The results showed that the LDA topic model could realize the efficient extraction and deep mining of ATC operation risks through the operation data,and 7 topics were digged out,including the ATC human factors,special situation disposal,ground and air coordination,and so on.Among them,the topic of ATC human factors was the core topic,and it had strong correlation with other topics.The combination of Word2Vec and semantic network could more accurately mine the relationship between the risks,determine the priority of topics and identify the key risks.