摘要
为了有效解决具有非线性特征的水文预报精准度的问题,通过对反向传播BP神经网络的学习和研究,分析了变量间的相互信息,提出了系统间相关信息熵的概念,并建立了适合水文预测的自迭代反向传播神经网络模型。该模型通过对迭代因子的及时修正,在反向传播中不断调整网络的权值和阈值,从而在很大程度上改善了传统BP算法所带来的不足,提高了预测的精度。实际的应用研究表明,自迭代反向传播模型的预测效果优于传统预测模型。
To effectively solve the accuracy problem of the hydrological forecasting with non-linear characteristics, by learning and researching the back-propagation BP neural network, analyzing the mutual information between the variables, the concept of the related information entropy between the systems is proposed, and a self-iterative back-propagation neural network model suitable for hydrological forecast is set up. This model improved the deficiency of the traditional BP algorithm greatly by correcting the iterative factor timely, adjusting the weights and thresholds of the network constantly in the back-propagation progress. Finally, this method increased the forecast accuracy. In the applied research, the forecast result of the self-iterative back-propagation model is better than the traditional one.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第2期398-400,405,共4页
Computer Engineering and Design
基金
"十一五"国家科技支撑计划重点基金项目(2006BAB04A13)
关键词
数据挖掘
神经网络
反馈输入
自迭代反向传播
相关信息熵
data mining
neural network
feedback input
self-iterative back-propagation
information entropy
