基于弹性小世界回声状态网络的非线性时间序列预测  

RESEARCHES ON NONLINEAR TIME SERIES PREDICTIONUSING ELASTIC SMALL WORLD ECHO STATE NETWORK

作　　者：任条娟 钟陈健[1] 刘半藤 郑启航 Ren Tiaojuan;Zhong Chenjian;Liu Banteng;Zheng Qihang(School of Information Science&Engineering,Changzhou University,Changzhou 213164,Jiangsu,China;College of Information Science and Technology,Zhejiang Shuren University,Hangzhou 310015,Zhejiang,China)

机构地区：[1]常州大学信息科学与工程学院,江苏常州213164 [2]浙江树人大学信息科技学院,浙江杭州310015

出　　处：《计算机应用与软件》2021年第6期256-261,共6页Computer Applications and Software

基　　金：浙江省公益技术研究项目(LGF18F010005)。

摘　　要：非线性时间序列预测在物理、医学和经济等领域具有广泛的需求。为提高非线性时间序列的预测精度,提出一种基于弹性小世界回声状态网络的非线性时间序列预测方法。采用小世界网络作为储备池,提高模型适应性;构建l 1范数和l 2范数联合的弹性惩罚模型,提高模型求解的数值稳定性和精确性。对非线性时间序列进行仿真实验,该方法相比常规方法具有更好的稳定性和更高的预测精度。Nonlinear time series prediction is widely used in the fields of physics,medicine and economy,etc.Therefore,a nonlinear time series prediction algorithm based on elastic small world echo state network is proposed in order to improve the accuracy of prediction.The Small world network was used as a reservoir to improve the model adaptability;the elastic penalty model consists of l 1 norm and l 2 norm was constructed to improve numerical stability and accuracy of model solving.Finally,the nonlinear time series was simulated.Compared with the other methods,the method has better stability and higher prediction accuracy.

关 键 词：小世界 弹性惩罚 非线性 预测方法 预测精度 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]