欢迎访问林业科学,今天是

林业科学 ›› 2008, Vol. 44 ›› Issue (1): 124-127.doi: 10.11707/j.1001-7488.20080120

• 论文 • 上一篇    下一篇

基于过程神经网络的木材生长轮密度预测

葛利1 陈广胜2   

  1. 1.哈尔滨商业大学计算机与信息工程学院,哈尔滨150056;2.东北林业大学,哈尔滨150040
  • 收稿日期:2007-07-11 修回日期:1900-01-01 出版日期:2008-01-25 发布日期:2008-01-25
  • 通讯作者: 陈广胜

Timber Growth Ring Density Forecast Based on Process Neural Network with Time-Varied Input and Output Functions

Ge Li1,Chen Guangsheng2   

  1. 1.College of Computer and Information Engineering,Harbin University of Commerce Harbin 150056;2.Northeast Forestry University Harbin 150040
  • Received:2007-07-11 Revised:1900-01-01 Online:2008-01-25 Published:2008-01-25

摘要: 提出一种基于过程神经网络的木材生长轮密度长期预测方法。本方法利用输入输出均为时变函数的过程神经网络输出为时变函数的特点,将原始数据拟合为输入函数并表示为一组正交基的展开形式后,使用混合遗传算法训练过程神经网络,得到过程神经网络的输出函数,以此实现木材生长轮密度的一次多步长期预测,通过与传统时间序列预测方法比较,预测精度得到显著提高,并为时间序列长期预测问题提供新方法。

关键词: 生长轮密度, 长期预测, 混合遗传算法, 过程神经网络

Abstract: A long-term forecast method of timber growth ring density based on process neural network was proposed in this paper.Making use of the feature of process neural network with output function,after raw data are fitted to input functions and are represented as expansion of a same orthogonal basis,process neural networks is learned by hybrid genetic algorithm and the output function is obtained.The multi-pace long-term forecast is once achieved.Comparing with tradition time-series forecast method,the forecast precision is apparently improved.And a new method of time series long-term forecast question is provided in this paper.

Key words: growth ring density, long-term forecast, hybrid genetic algorithm, process neural network