关键词: 暴雨洪涝灾害; 危险性分析; 云模型; 支持向量机; 遗传算法;

Abstract: To accurately assess the impact of rainstorm events on human society, it is significant to consider the uncertainties of the rainstorm sequence.The assessment on urban rainstorm hazards by cloud model can better qualify the uncertainty and is regarded as a more effective assessment tool.The historical series of four kinds of disaster-inducing factors of rainstorm(2002～2012) of Shenzhen City were selected as the evaluation index, and the weight of each factor was calculated by using the combined weighting method, then the assessment results on historical rainstorm hazard were obtained through the cloud model.Support vector regression model based on genetic algorithm parameter optimization(GA-SVR) was used to predict the values of four disaster-inducing factors from 2013 to 2016.The GA-SVR combined with the cloud model(GA-SVR-C) was used to predict the hazard levels in the predicted year.Meanwhile, the prediction results from the SVR and BP-network model were used for comparative analysis.It was proved that the GA-SVR-C model performed better than the other two models in terms of the prediction accuracy.The hazard assessment results of GA-SVR-C model are basically consistent with the real facts, which can reflect the hazard levels of urban rainstorm disaster well.

Key words: rainstorm and flood disaster; hazard analysis; cloud model; support vector machine; genetic algorithm;