摘要
介绍了逐步回归、岭回归、偏最小二乘回归、RBF神经网络、主成分RBF组合模型的基本思路与特点。以陈村大坝变形计算为例,分别建立了各种回归模型,比较了各种模型的优缺点,指出线性统计模型中偏最小二乘回归法的拟合精度及解释能力优于逐步回归、岭回归法;RBF神经网络、主成分RBF组合模型优于线性统计模型,主成分RBF组合模型最优,拟合及预测精度最好。
Article describes characteristics and the basic ideas of the stepwise regression, ridge regression, partial least- squares regression,RBF neural networks, principal components RBF combined model. To Chencun dam deformation, for example,a variety of regression models were established to compare the advantages and disadvantages of various models and pointed out that in a linear statistical model, partial least squares regression fitting precision and explanatory are power than in stepwise regression, ridge regression. RBF neural network, principal component RBF combined model is superior to linear statistical model. Principal component model for the optimal combination of RBF has the best fitting and forecasting accuracy.
出处
《水电能源科学》
北大核心
2009年第5期77-80,共4页
Water Resources and Power
基金
国家自然科学基金资助重点项目(50809025
50539010)
国家自然科学基金资助项目(50879024)
关键词
因子相关性
统计回归模型
主成分分析
RBF神经网络模型
factor correlativity
statistical regression model
principal component analysis
RBF neural network model
作者简介
许后磊（1987-）．男．硕士研究生，研究方向为大坝安全监测．E—mail：xuhoulei1987@163．com
