中国物理B ›› 2003, Vol. 12 ›› Issue (6): 594-598.doi: 10.1088/1009-1963/12/6/304

• • 上一篇    下一篇

Determining the input dimension of a neural network for nonlinear time series prediction

刘红星1, 高敦堂1, 都思丹1, 张胜2   

  1. (1)Department of Electronic Science and Engineering, Nanjing University, Nanjing 210093, China; (2)Department of Electronic Science and Engineering, Nanjing University, Nanjing 210093, China; Department of Physics, Nanjing Normal University, Nanjing 210097, China
  • 收稿日期:2002-11-18 修回日期:2003-02-09 出版日期:2003-06-16 发布日期:2005-03-16
  • 基金资助:
    Project supported by the National Natural Science Foundation of China (Grant Nos 59905011 and 60275041).

Determining the input dimension of a neural network for nonlinear time series prediction

Zhang Sheng (张胜)ab, Liu Hong-Xing (刘红星)a, Gao Dun-Tang (高敦堂)a, Du Si-Dan (都思丹)a   

  1. a Department of Electronic Science and Engineering, Nanjing University, Nanjing 210093, China; b Department of Physics, Nanjing Normal University, Nanjing 210097, China
  • Received:2002-11-18 Revised:2003-02-09 Online:2003-06-16 Published:2005-03-16
  • Supported by:
    Project supported by the National Natural Science Foundation of China (Grant Nos 59905011 and 60275041).

PDF (PC)

592

赞

摘要： Determining the input dimension of a feed-forward neural network for nonlinear time series prediction plays an important role in the modelling. The paper first summarizes the current methods for determining the input dimension of the neural network. Then inspired by the fact that the correlation dimension of a nonlinear dynamic system is the most important feature of it, the paper presents a new idea that the input dimension of the neural network for nonlinear time series prediction can be taken as an integer just greater than or equal to the correlation dimension. Finally, some validation examples and results are given.

Abstract: Determining the input dimension of a feed-forward neural network for nonlinear time series prediction plays an important role in the modelling. The paper first summarizes the current methods for determining the input dimension of the neural network. Then inspired by the fact that the correlation dimension of a nonlinear dynamic system is the most important feature of it, the paper presents a new idea that the input dimension of the neural network for nonlinear time series prediction can be taken as an integer just greater than or equal to the correlation dimension. Finally, some validation examples and results are given.

Key words: nonlinear time series, prediction, phase space reconstruction, neural network, input dimension

中图分类号:  (Time series analysis)

  • 05.45.Tp
07.05.Mh (Neural networks, fuzzy logic, artificial intelligence)