基于潜变量回归模型的城市火灾风险评估  

Risk Evaluation of Urban Fire by Regression Model with Latent Variables

作　　者：丁倩[1] 尚涛[1] 刘天桢[1] 

机构地区：[1]武汉大学城市设计学院,湖北武汉430072

出　　处：《中国安全科学学报》2012年第7期159-163,共5页China Safety Science Journal

基　　金：国家自然科学基金资助(51008238)

摘　　要：为评估城市火灾风险等级,针对城市火灾发生的可能性与严重性2个基本问题,建立评估指标体系,提出一种基于潜变量回归的评估模型,并给出无须迭代的确定性算法。该模型和算法无需事先指定风险等级评估分数的训练样本,不必事先指定指标汇总的加权系数,利用合理的配方约束,完成潜变量回归,实现城市火灾风险按地区或者按行业的等级评估,利用现成的软件计算,同样适用于其他具有类似数据结构的评估问题。In order to evaluate the grades of urban fire risk, an index system was built according to two basic problems： the fire risk possibility and its seriousness, an evaluation model based on regression with latent variables was built, and a definite algorithm without iteration was given. The model and algorithm do not need training sample with risk grade and appointed weighted coefficients for summary beforehand, and realize regression with latent variable by use of reasonable prescription constraint. The model and algorithm can realize evaluation of urban fire risk by areas or industries with convenient computation software, and are also suitable to other evaluation problems with similar data constructions.

关 键 词：城市火灾 风险评估 指标体系 潜变量回归 确定性算法 配方约束 

分 类 号：X92[环境科学与工程—安全科学]