摘要
针对最小二乘支持向量机(LS-SVM)参数不易确定的问题,利用遗传神经网络模拟LS-SVM计算结果与参数之间的关系,提出了一种基于遗传神经网络(GA-BP)的参数选择方法,该方法利用正交分解法构建训练参数组,并将参数组代入最小二乘支持向量机以获得计算输出值,然后将计算输出值与训练参数组代入遗传神经网络进行训练并获得合适的LS-SVM参数。最后以土石坝渗流分析为例进行验证,结果表明该方法对优化选择最小二乘支持向量机参数十分有效,预测精度可达10-4。
The parameters in Least Squares Support Vector Machines (LS-SVM) are not easy to determine. A new parameter-selecting method that simulates the relation of LS-SVM parameters with their results through a genetic neural network is put forward. The training parameter group is constructed with proper orthogonal decomposition. The resulting output is obtained by putting the training parameter group into LS-SVM. By substitution of the output and the train parameters set into the genetic neural network, the most suitable parameters of LS-SVM are determined. Finally, the example of earth dam seepage is considered. Simulation results show that this method can yield optimized parameters and that the prediction accuracy is over 10^-4.
出处
《水利水电科技进展》
CSCD
北大核心
2009年第1期8-11,共4页
Advances in Science and Technology of Water Resources
基金
河南省重大公益性科研项目(081100911200)
关键词
最小二乘支持向量机
参数反演
遗传神经网络
Least squares support vector machines
back analysis
genetic neural network