基于WNN-RAGA的非线性组合预测方法  

Nonlinear combination forecasting approach based on WNN-RAGA

作　　者：王硕[1,2] 杨善林[2] 马溪骏[2] 

机构地区：[1]合肥工业大学人文经济学院,合肥230009 [2]合肥工业大学计算机网络系统研究所,合肥230009

出　　处：《系统工程理论与实践》2008年第12期160-165,共6页Systems Engineering-Theory & Practice

基　　金：国家自然科学重点基金(70631003);国家自然基金重大研究计划(90718037);教育部人文社会科学规划项目(07JA790109);安徽省高等学校省级教学研究项目(2005126)

摘　　要：将小波变换与神经网络结合构成WNN非线性组合模型,弥补单一预测方法的不足,利用RAGA的全局优化能力,优化WNN学习的网络参数,解决传统WNN网络学习算法在学习后期收敛速度慢、存在局部最小值以及训练结果不稳定的问题.实例的预测结果显示出这种基于WNN-RAGA非线性组合预测的良好预测性能.Building WNN nonlinear combination models by combining wavelet transform with neural network to supply a gap of single forecasting method, optimizing network parameters of WNN study by RAGA＇ s abilities of whole optimization to resolve the problems which are appeared in the late phase of WNN network study algorithm. These problems contains that convergence rate is low, local least value is existed and training results are instable. The forecasting results of instance reveals a favorable forecasting capability based on WNN-RAGA nonlinear combination forecasting.

关 键 词：小波神经网络 加速遗传算法 非线性组合预测 预测精度 

分 类 号：G304[文化科学]