基于免疫遗传算法的GMDH网络模型及其应用  

GMDH network model and its application based on immune genetic algorithm

作　　者：陈洪[1,2] 陈森发[1] 

机构地区：[1]东南大学系统工程研究所,江苏南京210096 [2]南京工程学院数学研究所,江苏南京211167

出　　处：《江苏大学学报（自然科学版）》2009年第6期610-613,共4页Journal of Jiangsu University：Natural Science Edition

基　　金：教育部博士点基金资助项目(20060286005);江苏省高校自然科学基金资助项目(07KJD580085)

摘　　要：针对传统GMDH网络建模用最小二乘法辨识参数时常常陷入局部极小导致模型预测效果不理想的问题,提出将免疫算法与遗传算法结合起来,引入到GMDH网络,来辨识其部分描述式系数.给出自适应免疫遗传算法,构建了基于该算法的GMDH网络模型,并将IGA-GMDH模型应用于苏州一交叉口的交通流量数据的仿真研究.结果表明,该算法既保证了全局寻优和所求解的精度,又进一步提高了全局与局部寻优能力;所构建的IGA-GMDH网络模型比传统的GMDH网络预测精度高.Considering the problem in modeling of the conventional GMDH network,the least square me-thod used to identify coefficients often led to local minimization,and forecasting was not inaccurate enough,a concept was proposed by combining the immune algorithm and genetic algorithm into the GMDH network to identify its partial descriptive coefficients. An adaptive immune genetic algorithm was provided,and based on this algorithm an IGA-GMDH network model was constructed. This model was applied to simulate the traffic volume at an intersection of Suzhou. The results show that the model can secure global optimization and high precision of corresponding solutions,improve the capacities of global optimization and local optimization,and possesses higher precision of forecasting than the conventional GMDH network.

关 键 词：GMDH网络 遗传算法 免疫算法 模型 预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] U491.14[自动化与计算机技术—控制科学与工程]