基于灰关联度的事故聚类分析

Cluster Analysis of Accidents Based on Degree of Gray Correlation

事故是来自系统和环境的风险 ,危害人类的生产、生活和生存活动。对事故进行科学分类是安全管理和事故处理的重要基础。笔者针对现行事故等级分类标准的不足 ,分析了事故造成的人员伤亡和经济损失等多属性特征 ;通过灰关联度确定了各属性指标的不同权重 ,定义了事故间的加权欧氏距离 ;根据事故属性指标的相似度 ,利用聚类分析对事故进行了分类 ;以 2 0 0 3年全国发生的 14起特别重大事故为例进行了实证分析。提出的事故分类思想和方法 ,综合考虑了事故所造成的损失 ,理论严谨、方法实用有效 ,可作为政府有关部门进行事故分类和事故处理的决策依据。

Accidents come from the risk of the system and environment, which do harms to production, life and existence of mankind. To categorize accidents scientifically is the important basis for safety management and dealing with the aftermath of accidents. In view of lack of gradation criteria for accidents, multiple characteristic attributes such as casualty and economic loss caused by accidents are analyzed. Different weight of each attribute is determined by the degree of gray correlation, by which the weighted Euclid distance among different accidents is defined. According to the similarity of accidental attributes, accidents are categorized with cluster analysis. Taking 14 especially severe accidents occurred in 2003 as examples, the analysis is exemplified, from which the ideology and method of accident categorization are derived. In this way, the loss caused by accidents is comprehensively taken into account with strong theoretical basis. The method is feasible and effective and could be used as the criteria in decision-making of relevant departments of the government in categorizing the accidents and dealing with the aftermath of accidents.