船舶溢油事故报告关键影响因子与行为偏好分析  

Analysis of the key influential factors and behavior tendencies of the oil-spill accident reports of ships

作　　者：王立坤[1] 

机构地区：[1]上海海事大学交通运输学院,上海201306

出　　处：《安全与环境学报》2013年第6期212-217,共6页Journal of Safety and Environment

基　　金：国家自然科学基金项目(70673060);教育部人文社会科学研究青年基金项目(12YJC630205)

摘　　要：为分析船舶溢油事故报告行为的影响因子和偏好,构建以环境因子和事故船方主体特征为自变量、报告选择为因变量的关系结构,从连通性、效率和效果3个角度完成了相关信息采集。基于调查问卷数据,分别利用Probit模型和贝叶斯网推理明确了影响事故报告行为的关键因子,基于ROC值对两个模型的预测精确度进行了评估,选择精确度较高的模型预测事故报告行为偏好。结果表明,在船工龄、职务、培训情况、溢油数量和事故时间会对溢油事故报告的具体行为偏好产生影响。In order to analyze the accident report behaviors of the oil spill, this article has established an influential relationship model with the environmental factors and the accidentprone ship party＇s charac teristics as the independent variables, and with the report behaviors as the dependent variables. The environmental factors include the site of the accident, the volume of the oil spills, the cause of the acci dent, the timejuncture of the oilspill accident, the size of the wind power and the shiptype. The accident ship party＇s characteristics in clude the onboard working time, duty and training experience. The accident report behaviors include the object party that the accident ship party has reported to, the time when the ship party reported, the way for measuring the volume of spilled oil and the contents of the ac cidentship in the report. Based on the questionnaire distribution and data collection from the experienced crew, we have first of all chosen a more accurate probit model on whether there is a constant to be tak en into account based on AIC value, BIC value and Hosmer Lemeshow chisquared value. Secondly, we have adopted the Probit model with Zvalue and Bayesian network model with pvalue as the principal devices for calculating the key influential factors involving the accident report behavior separately. And, thirdly, based on the ROC index, we have chosen a more suitable model from the probit and BN to forecast the accident behavior probability. The results of our trial use of our model indicate that the onboard working time, duty and training experience have important effects on the object party reported or to be reported. The onboard working time and the train ing experience are likely to have obvious effects on the report time. The volume of spilled oil and the time when the oilspill accident take place may all influence the manner for measuring the volume of the oil spilled. It is from the above said influential factors that we would like to suggest that efficient training helps the ship par

关 键 词：安全管理工程 船舶溢油应急处置 Probit模型 朴素贝叶斯网 事故报告行为偏好 

分 类 号：X951[环境科学与工程—安全科学]