大型商业综合体作为一种公共聚集场所，结构与功能复杂，人员密集，流动性大，极易发生火灾风危险。我国目前的火灾风险性评价工作主要依据现行防火规范标准，即强调每项指标达标。但大型商业综合体中存在一些规范无法规定的新型设计理念及结构布局，这对以现行规范为参考依据的火灾风险性评价工作提出巨大挑战。 本文针对大型商业综合体火灾事故中存在的突出问题以及火灾风险性评价中突显的矛盾，提出大型商业综合体火灾风险性评价体系的研究课题。主要做了以下几个方面的研究： 第一，火灾风险性评价体系的构建及指标因素的衡量分析。 （1）针对大型商业综合体的火灾发展及防火设计特点，构建了三级火灾风险性评价指标体系。 （2）针对超规范大型商业综合体的防火设计方法与现行防火规范之间的矛盾，将三级指标进一步划分为超规范设计类、符合规范设计类和严于规范设计类三大类别，并分析总结了火灾风险性评价工作中存在的问题。 （3）在上述分析的基础上，对火灾风险性评价指标进行衡量分析。 第二，提出了综合运用模糊层次分析法、模糊综合评价法和遗传神经网络的方法对大型商业综合体火灾风险性进行评价的设计构想。 （1）运用FAHP-FCE评价模型生成GA-BP神经网络的样本集。运用MATLAB编程，完成了模糊层次分析法对指标权重的计算过程；并运用模糊综合评价法分析得出各个样本的火灾风险等级。 （2）运用遗传算法优化BP神经网络的权值和阈值，构建了基于遗传神经网络的评价模型。运用MATLAB的神经网络工具箱和编写GA-BP程序，完成对GA-BP网络评价模型的训练、学习与测试。 第三，将训练好的网络应用于实例评价，结果表明：该评价模型在大型商业综合体火灾风险性评价中具有良好的适用性。

As public areas, because of its complication of function structure，personnel-intensive and flow of people, the large-scale Commercial Complex is easy to fire risk. The current fire risk assessment work are based on the current national code for fire protection, is put to use by the majority of the projects in our country. Nevertheless, the new design and the new distribution formats turn up in commercial buildings, all of which are gradually hard to be stipulated by some standards. According to current prominent problems existing in large-scale Commercial Complex’ accidents and deep contradictions shown in the fire risk assessment, we propose to do the research on the evaluation system in large-scale Commercial Complex .Basically has the following several aspects: Firstly, building fire risk evaluation system and measuring analysis the index factor. （1）According to the fire development and fire prevention design characteristics of large-scale commercial complex, we establish the three-level fire risk evaluation index system. （2）In view of the contradiction of the super standard large-scale commercial complex way of fire protection design and the current code for fire protection. We will three-level indicators are further divided into super specification design, conform to the specifications design and strict in specification design class three categories and analyzed the problems of the fire risk evaluation. （3）On the basis of above analysis, we measure analysis of the fire risk evaluation index. Secondly, it puts forward a design idea that use the FAHP-FCEM and GA-BP to make the comprehensive evaluation of the large-scale Commercial Complex’ fire risk. （1）Using the FAHP-FCEM to generate the sample set of the GA-BP neural network. Implement the process of calculating weight by Writing Fuzzy analytic hierarchy process (ahp) program. And using the method of fuzzy comprehensive evaluation to analysis fire risk grade of each sample . （2）By using genetic algorithm to optimize the weights and thresholds of BP neural network, the evaluation model of GA-BP network are respectively trained. And it is learned and tested by using the neural network toolbox of MATLAB and writing GA-BP program. Thirdly, the trained network was applied to the instance, and the applying results show that the evaluation model has good practicality in the fire risk assessment of the large-scale Commercial Complex．