金龙,况雪源,黄海洪,覃志年,王业宏. 2004. 人工神经网络预报模型的过拟合研究[J]. 气象学报, 62(1):62-70, doi:10.11676/qxxb2004.007
人工神经网络预报模型的过拟合研究
STUDY ON THE OVERFITTING OF THE ARTIFICIAL NEURAL NETWORK FOR ECASTING MODEL
投稿时间:2003-02-13  修订日期:2003-04-25
DOI:10.11676/qxxb2004.007
中文关键词:  神经网络  泛化性能  过拟合现象  预报建模
英文关键词:Artificial neural network  Generalization capability  Overfitting  Establishment of a forecasting model
基金项目:国家自然科学基金项目(40075021)
作者单位
金龙 广西壮族自治区气象减灾研究所, 南宁, 530022 
况雪源 广西壮族自治区气候中心, 南宁, 530022 
黄海洪 广西壮族自治区气象减灾研究所, 南宁, 530022 
覃志年 广西壮族自治区气候中心, 南宁, 530022 
王业宏 南京气象学院, 南京, 210044 
摘要点击次数: 2037
全文下载次数: 4212
中文摘要:
      针对神经网络方法在预报建模中存在的“过拟合”(overfitting)现象和提高泛化性能(generalization capability)问题,提出了采用主成分分析构造神经网络低维学习矩阵的预报建模方法。研究结果表明,这种新的神经网络预报建模方法,通过浓缩预报信息,降维去噪,使得神经网络的预报建模不需要进行适宜隐节点数的最优网络结构试验,没有“过拟合”现象,并且与传统的神经网络预报建模方法及逐步回归预报模型相比泛化能力有显著提高。
英文摘要:
      With the application of the artificial neural network(ANN) in the field of Atmospheric Science,a "bottle-neck" was found while the artificial neural network model was applied for weather forecast:the fitting precision of training sample could not be definitely determined to make the model showing its best forecasting capability.It was a key problem to be solved on the overfitting and generation capability of the ANN application theory area.Study on this problemis necessary for the further operating application of ANN in the field of Atmospheric Science.A new forecasting model has been proposed for model establishment by means of making a low-dimension ANN learing matrix through principal component analysis(PCA-ANN).The monthly rainf all of June、July and August were forecasted by using PCA-ANN,R-ANN(regression artificial neural network model) and SR(stepwise regression model) respectively.
HTML   查看全文   查看/发表评论  下载PDF阅读器
分享按钮