基于事故调查报告数据的主体责任文本挖掘及培训应用

基于事故调查报告数据的主体责任文本挖掘及培训应用摘要近年来，我国安全事故总体稳定，但安全事故时有发生，地方党政领导干部作为安全工作的“关键少数”仍然存在安全能力意识不足的情况，对此，需要加强安全教育培训，明确自身安全职责，加强自身安全工作能力。以往的安全教育培训停留于对单一事故的简单描述，或机械式地将几个事故汇集，缺少针对性与系统性，故本文采用文本挖掘的方法，结合T℉-DF算法对100篇重特大事故调查报告进行分析，经过一系列的文本处理，形成了高频责任文本关键词，通过词云图进行可视化数据展示，并建立T℉-IDF矩阵，对省市区街道四个级别的高频责任差异进行分析。进一步将责任文本关键词进行分类整理，总结出常见的事故责任类别：落实指示精神、落实政策、落实工作部署、隐患排查治理、事故教训吸取、安全指标设立、违规审批、形式主义、对下级问题失察失管。针对目前地方党政领导干部安全教育的不足，提出了模块化教学的方案，明确模块化教学的优势，从目标、内容、教育方式三个角度对安全教育培训模块进行了设计，并将总结出的责任类型分模块融入到安全教育内容中，使得地方党政领导干部安全教育培训加强了针对性、系统性。随着安全教育培训的进一步发展完善，期望为后续安全教育研究以及培训工作提供一些有效的参考。关键词文本挖掘：TF-DF算法：安全教育培训：词云图-I-Text Mining and Training Using of SubjectiveResponsibility Based on Accident InvestigationReport DataAbstractIn recent years,China's overall stability of security accidents,but securityaccidents occur from time to time,the local party and government leaders andcadres as a "key minority"of security work still exists in the lack of awarenessof security capabilities,which requires the strengthening of security educationand training,clear their own security responsibilities,and strengthen their ownsecurity work capacity.Previously,safety education and training stayed in thesimple description of a single accident,or mechanically bring together severalaccidents,lack of relevance and systematic,so the text using text miningmethods,combined with the TF-IDF algorithm to analyze the 100 investigationreports of serious accidents,after a series of text processing,the formation ofhigh-frequency responsibility text keywords,through the word cloud map tovisualize the data display,and the establishment of the TF-IDF matrix.TF-IDFmatrix to analyze the high-frequency responsibility differences at four levels ofprovinces,municipalities and streets.The responsibility text keywords arefurther categorized and organized to summarize the common accidentresponsibility categories:implementation of the spirit of instructions,implementation of policies,implementation of work deployment,hidden dangerinvestigation and management,lessons learned from accidents,theestablishment of safety indicators,violation of approvals,formalism,and failureto monitor and control the problems of subordinates.-Ⅱ-In view of the current shortcomings of the safety education of local party andgovernment leaders and cadres,a modularized teaching program is proposed tomake clear the advantages of modularized teaching,and the safety educationand training modules are designed from three perspectives of objectives,content,and education methods,and the summarized types of responsibilities areintegrated into the content of safety education in modules,which makes thesafety education and training of local party and government leaders and cadresto strengthen their relevance and systematicity.With the further developmentand improvement of safety education and training,it is expected to providesome effective references for subsequent safety education research as well astraining work.Keywords Text mining,TF-IDF algorithm,Safety education and training,W