基于数据挖掘的化工生产事故致因主题抽取

为充分挖掘化工生产事故数据中的有效信息和潜在规律，提高对化工事故认知水平，针对某化工集团2010—2016年共1 578起事故数据，利用社会网络分析等方法揭示事故要素间的关联关系；运用潜在狄利克雷分配（LDA）模型进行事故聚类，并抽取到5个事故致因主题。研究结果表明:LDA主题模型等数据挖掘技术能有效挖掘大量事故数据中的潜在信息；5个事故致因主题中，4个涉及到人因或组织层面的缺陷；员工注意力不集中和现场风险管理不足这2个致因主题间具有较强相关性；员工注意力不集中、现场风险管理不足以及设备问题是导致事故发生的主要原因。

Topic extraction on causes of chemical production accidents based on data mining

In order to fully exploit the effective information and potential laws in the data of chemical production accidents,and improve the cognitive level of chemical accidents,according to the data of 1 578 accidents in a chemical industry group from 2010 to 2016,the association relationship between the accident elements was revealed by means of social network analysis.The Latent Dirichlet Allocation (LDA) model was applied to conduct the accident clustering,and five topics of accident causes were extracted.The results showed that the data mining techniques such as LDA topic model could effectively mine the potential information in a large amount of accident data.Among the five topics of accident causes,four topics involved the human factors or organizational defects.Two topics of accident causes including the lack of concentration of employees and the inadequate on site risk management had a strong correlation.The lack of concentration of employees,the inadequate on site risk management,and the equipment problems were the main causes of accidents.