安全、健康和环境2026，Vol.26Issue(6)：82-87,6.DOI:10.3969/j.issn.1672-7932.2026.06.011
炼化企业安全隐患分析中LDA模型参数敏感性研究
Sensitivity Study on Parameters of LDA Model in Safety Hazard Analysis of Refining and Chemical Enterprises
摘要
Abstract
Focusing on the unstructured hazard data within the digital systems of the dual prevention mecha-nism in refining and chemical enterprises,this study proposes a hazard analysis method based on text mining and the Latent Dirichlet Allocation(LDA)topic model.A total of 10 555 hazard records collected from a Petro-chemical company between 2022 and 2023 were utilized as the research sample.To enhance the accuracy of word segmentation,a specialized safety lexicon was constructed.Subsequently,the Term Frequency-Inverse Document Frequency(TF-IDF)algorithm was applied to extract core keywords,and a random search method was employed to optimize the parameters of the LDA model.Consequently,five fundamental hazard topics and three detailed sub-topics were successfully extracted.The results indicated that this method can effectively re-veal the characteristics of high-frequency hazards.Therefore,these findings provide robust technical support for enterprises to precisely formulate targeted hazard inspection strategies.关键词
炼化企业/双重预防机制/文本挖掘/LDA主题模型/隐患分析/参数优化Key words
refining and chemical enterprises/dual prevention mechanism/text mining/Latent Dirichlet Allocation(LDA)topic model/hidden hazard analysis/parameter optimization分类
资源环境引用本文复制引用
厉建祥,徐杨,尹志攀,李千登..炼化企业安全隐患分析中LDA模型参数敏感性研究[J].安全、健康和环境,2026,26(6):82-87,6.基金项目
中国石化化工事业部项目(CHG26007),流程驱动的炼化基层岗位HSE职责落实与评价技术研究. （CHG26007）
