基于自适应遗传算法和灰色关联度的摆动序列建模研究

灰色摆动序列经过动态指数变换具有灰指数特性,对变换后数据序列建立GM(1,1)可以进一步提高预测精度.为了确定动态指数变换函数的最优值,提出利用具有全局搜索能力的自适应遗传算法对该参数进行辨识.首先用指数变换后序列和最优参考指数序列的灰色关联度构造适应度函数,然后用自适应遗传算法求解待辨识参数的非劣解,最后引入辨识参数进行序列变换,并对变换后序列建立GM(1,1).自适应交叉和变异概率可以控制非劣解替换和保持种群多样性, 所以该变换方法使摆动数列具有更好的灰指数规律,可以大大提高摆动序列的预测精度.应用实例结果表明该方法的有效性.

Modeling Research of Grey Wobbly Sequence Based on Adaptive Genetic Algorithm and Grey Relation Degree

In this paper,grey wobbly sequence takes on grey exponent character by an especial dynamic exponent transformation.After that,if GM(1,1) for the transformed sequence is set up,the forecasting accuracy is improved more.And in order to get optimized parameters of dynamical exponent transforming function,adaptive genetic algorithm that has whole searching ability is introduced.First,the grey relation degree between an exponent transforming sequence and a best referenced exponent sequence serves as a fitness function,and the parameters are solved by Genetic Algorithm.Then the grey wobbly sequence is transformed by the dynamical exponent transforming function.Finally,GM(1,1) is set up for transformed sequence.The variety of population is kept by means of(adaptive) probability of crossover and mutation.Simulation examples show the effectiveness of the proposed approach,so the method makes the grey wobbly sequence adapt to exponent law better,and the forecasting accuracy is advanced greatly.Simulation of examples shows the effectiveness of the proposed approach.