一种基于最优化的极值分布建模方法

为提高极值分布中样本数据序列分布拟合精度,对改进的综合模型采用遗传算法实现分布参数的寻优。建立了灾难性事件的样本极值分布数学模型。采用非线性回归方法导出样本极值与累积概率之间的映射关系,考虑样本极值的上限和拟合的误差,建立了极值分布的综合模型。采用改进遗传算法,将模拟退火算法应用到遗传算法中,以模型误差为目标函数进行优化,从而确定函数模型中的分布参数,实现了拟合精度的提高。

Approach of extreme value distribution modeling based on optimization

To improve the fitting precision of simple sequence in extreme value distribution,the improved integration model's parameters are optimized by genetic algorithms(GA).The mathematic model of the disasters' extreme value distribution is established,and the mapped relationship between accumulation probability and extreme simples is developed by nonlinear regress.Considering the upper limit of the extreme value and the fitting error,the integration distribution model is established.Simulated annealing(SA) is applied to GA,and the model error is regard as an object function.Therefore,the distribution parameter in integration evaluation model is ascertained by optimization,and the fitting precision is enhanced.