基于蚁群算法改进的扩散函数类内集-外集模型及其灾害评估实验

Interior-outer-set models based on diffusion function improved by ant colony algorithm and their experiments on disaster risk assessment

针对自然灾害的复杂性、不确定性以及灾害评估信息的不完备等问题,基于内集-外集模型引入了小样本信息扩散原理以获取灾害事件与其发生概率之间的模糊关系。鉴于常规的扩散函数类的内集-外集模型中论域步长的选取是基于正态分布的,不仅要求样本容量较大,且有一定的主观性和应用局限性,为此引入蚁群聚类思想来划分样本,以寻求最优区间数;引入经验窗宽来简化扩散函数参数的选取;通过与原模型进行计算机仿真实验对比以及对广东省台风暴潮案例的风险评估,检验了所提出的改进内集-外集模型的可行性、实用性与有效性。为实际的自然灾害小样本事件风险评估建模拓展了新的思路和方法途径。

Aiming at the complexity,uncertainty of natural disasters and the imperfection of disaster assessment information,the interior-outer-set models based on information diffusion function(IOSMIDF) were introduced to obtain the fuzzy relation between disaster events and their probabilities of occurrence for small sample problems.However,the traditional IOSMIDF has some deficiencies such as:(1) the domain and the step length are chosen with some subjectivity from data only in normal distribution and abundant quantity;(2) the parameter selection of the diffusion function has certain complexity and ambiguity.Therefore,ant colony algorithm and empirical window width were introduced to reduce parameter selection of diffusion function,and the interior-outer-set model was developed.When analyzing the tremendous simulation results and the typhoon storm surge affected index of Guangdong Province from l949 to 2005,the risk evaluation model turned out to be in quite approximate disaster conformity with the actual situation.Thus,it can be concluded that the interior-outer-set model we have proposed can be a useful guide for typhoon storm surge and similar disaster risk assessment.