选择显著因子的BP神经网络模型研究  

Research of BP Neural Network Model with Selecting Significant Factor

作　　者：许孝臣[1] 戴春华[1] 邓成发[1] 

机构地区：[1]浙江省水利河口研究院,浙江杭州310020

出　　处：《人民黄河》2011年第12期121-122,125,共3页Yellow River

基　　金：浙江省属科研院所专项创新团队建设与人才培训项目(2009F20021)

摘　　要：为了更加准确有效地分析大坝安全监测资料,保证大坝的安全、有效运行,分析了神经网络和回归分析的优缺点,提出了选择显著因子的BP神经网络模型分析方法。实例分析结果表明:神经网络和回归分析具有较强的互补性,为了保证资料分析的可靠性,可同时运用这两种方法;选择显著因子的BP神经网络分析方法避免了因子之间的相关性导致预测精度不高的缺点,减小了神经网络过度训练的几率,在保持神经网络拟合精度的同时又具有较高的预测数度。In order to be more accurate and effective analysis of the dam safety monitoring data and ensure the safe and effective operation of the dam,both the advantages and disadvantages of neural network and regression analysis were analyzed, and BP neural network model with selecting significant factor was proposed. Conclusions are concluded by example： neural network and regression analysis have both advantages and disadvanta- ges, however they have strong complementarities, in order to ensure the reliability, both neural network and regression analysis should be used; the method of neural network model with selecting significant factor avoids the shortcomings of prediction accuracy not high because of the correlation between factors,reduces the chance of over-training neural network,that maintain the accuracy of the neural network fitting,also has a high predictive ability.

关 键 词：逐步回归分析 显著因子 BP神经网络 大坝监测 

分 类 号：TV697[水利工程—水利水电工程]