An Analytic Network Process Approach for Rapid Loss Assessment of High Casualty Fires in China

Mei Peng,Li Guohui,Lu Song,Zhang Heping
DOI: https://doi.org/10.1007/s10694-013-0329-6
IF: 3.605
2013-01-01
Fire Technology
Abstract:Focusing on the high casualty fires (HCFs), an analytic network process (ANP) approach was proposed for rapid loss assessment. The networks of situation index (SI) and consequence index (CI) were built. A dataset of 469 HCF cases in China from 2002 to 2010 was established. The relative weights of all attributes and the values of CI and SI were calculated and analyzed. The results show that the casualty factors should be emphasized in loss assessment by CI. The elements of day, hospital, and smoking are most likely to cause huge loss for the attributes of time, place, and cause in the SI network, respectively. The correct classification rates of accident level were compared between ANP, multinomial logistic regression, multi-layer perceptron network, and radial basis function network. It is indicated that ANP performs best for HCFs of the high and extreme level or when the influence of accident level is considered. The relationship between CI and SI for HCFs of the high and extreme level was obtained by the weighted least square based polynomial fitting and can be used to generate meaningful results of extreme possibility for decision making.
What problem does this paper attempt to address?