基于文本挖掘的建筑施工坍塌事故成因网络分析

Network Analysis of Construction Collapse Accidents Based on Text Mining

为查清建筑施工坍塌事故发生的特征,达到预防建筑施工坍塌事故和缓减事故伤害的目的。利用Python爬取安全管理网坍塌事故模块中2013—2021年200份建筑施工坍塌事故调查报告,并借助文本挖掘方法对调查报告进行分词处理、成因选择后,得到32个事故成因高频关键词。基于社会网络中心性分析、核心边缘分析、成因集合分析和凝聚子群分析,结合事故致因理论,划分事故成因,并利用事故成因网络图进行可视化。研究结果表明:(1)安全管理不到位,违反施工方案等7个核心成因具有地位高、联动强、路径短三大特征;(2)个别核心成因的个体网络中心度与度中心度相差较大;(3)存在违章作业,应急防范难以落实,承发包违规,模板失稳4类问题。

In order to find out the characteristics of construction collapse accidents,to achieve the purpose of preventing construction collapse accidents and reducing accident injuries.Python was used to crawl 200 investigation reports of collapse accidents in construction from 2013 to 2021 from the collapse accident module of safety management network.After word segmentation processing and cause selection of investigation reports by text mining method,32 high-frequency keywords of accident cause were obtained.Based on social network centrality analysis,core edge analysis,causal set analysis and cohesive subgroup analysis,combined with accident causation theory,the causes of accidents are divided,and the network diagram of accident causes is used for visualization.The results show that:(1)the seven core causes,such as inadequate safety management and violation of construction plan,have three characteristics:high status,strong linkage and short path;(2)the individual network centrality of individual core causes is significantly different from that of degree centrality;(3)there are four types of problems:illegal operation,difficult to implement emergency prevention,illegal contracting and template instability.