基于ISODATA的踩踏生成特征与集群风险关联性研究

Analysis of correlation between stampede generation characteristics and clustering risk using ISODATA algorithm

踩踏事故统计研究领域较少综合考虑事故致因因素与风险的关联程度,导致单因素统计结果无法准确阐明不同风险等级事故生成特征的问题,因此引入ISODATA算法开展研究。收集2000—2020年国内外215起踩踏事故,将描述性事件量化为特征向量,构建结构化数据集;采用ISODATA算法进行聚类迭代运算,得到7个集群,根据每个集群中不同等级事件的风险指数(R_(ki))确定高、低风险集群;通过多因素分布规律分析,阐明不同风险集群的生成特征。研究表明,ISODATA算法可显著识别踩踏事故多维特征与风险等级的关联,准确获得多因素与单因素统计的相似性和差异性,为踩踏预防管理提供依据。

The correlation degree between accident causal factors and risk is seldom considered in the field of stampede accident statistics research,which leads to the single factor statistical results that can not accurately clarify the characteristics of accidents with different risk levels.To solve the problem,the ISODATA algorithm is introduced to carry out the research.Firstly,this paper collected 215 human stampede accidents at home and abroad from 2000 to 2020.From the aspects of the accident site,accident type,risk grade,human factors,material factors,environment,management,etc.,this paper comprehensively analyzed the expression vector of the causal factors of stampede accident,quantified the descriptive events into feature vectors,and constructed a structured dataset.After that,this paper used the ISODATA algorithm to cluster iteratively and got 7 clusters.Cluster risk index(R_(ki))represented the possibility of a cluster in a certain degree of severity.It was constructed with the parameter that the proportion of events of different accident levels in each cluster and the proportion in the dataset.According to the R_(ki) of different levels of events in each cluster,it determined the high-risk and low-risk clusters.At last,the formation characteristics of different risk clusters were analyzed through the multi-factor distribution law.The results show that the proportion of high-risk clusters in temples and schools is 39%and 28%respectively,which is different from that of single factor stadiums(23%).The proportion of festival celebration in the accident type is 61%,which is similar to that of the single factor(26%).In addition,the multi-dimensional analysis on the characteristics of human factors,materials factors,environment,management,and other aspects of clusters with different risk levels is carried out,and the results are consistent with the reality.The ISODATA algorithm can significantly identify the correlation between the multi-dimensional characteristics and risk levels of stampede accidents,accurately obtain the similarity and difference between multi-factor and single factor statistics,and provide the basis for the prevention and management of human stampede.