摘要
针对使用传统极限学习机实现大坝变形预测中,因影响因子复杂导致隐藏层个数难以确定的问题,该文提出一种基于极限学习机与弹性网络支持下的大坝变形预测模型。采用极限学习机算法,将大坝变形影响因子由原本的空间映射到极限学习机的特征空间,建立影响因子与变形结果之间的非线性联系,同时将非线性模型转换成一个线性模式求解问题,并使用弹性网络求解该模型。对比基于极限学习机回归与最小二乘回归算法的实验表明:弹性网络拥有更好的稳定性,改善了极限学习机难以处理的过拟合现象,减弱了因训练集样本不同造成的预测误差大的影响,只需任意设置足够数量的隐含层神经元个数就能得到可靠的预测结果,简化了基于极限学习机的大坝变形预测面临的隐含层神经元个数取舍问题。
In order to realize the dam deformation prediction using the traditional extreme learning machine,the dam deformation prediction model based on the extreme learning machine and the elastic network is proposed because the number of hidden layers is difficult to determine due to the complexity of the influence factors.Using the extreme learning machine algorithm,the dam deformation influence factor is mapped from the original space to the feature space of the extreme learning machine,and the nonlinear relationship between the influence factor and the deformation result is established,and the nonlinear model is transformed into a linear mode solving problem.And use the elastic network to solve the model.Experiments based on extreme learning machine regression and least squares regression algorithm show that the elastic network has better stability,improves the over-fitting phenomenon faced by the extreme learning machine,and weakens the prediction error caused by different training set samples.The effect is that only a sufficient number of hidden layer neurons can be arbitrarily set to obtain reliable prediction results,which simplifies the problem of the number of hidden layer neurons faced by the dam deformation prediction based on the extreme learning machine.
作者
陈优良
肖钢
胡敏
黄劲松
CHEN Youliang;XIAO Gang;HU Min;HUANG Jinsong(School of Architectural and Surveying&Mapping Engineering,Jiangxi University of Science and Technology,Ganzhou,Jiangxi 341000,China;Zhejiang Intellectual Spectrum Engineering Technology Co.,Ltd.,Zhejiang 313000,China)
出处
《测绘科学》
CSCD
北大核心
2020年第11期20-27,40,共9页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41261093)
江西省教育厅科技项目(GJJ170522)。
关键词
变形预测
极限学习机
线性模式求解
弹性网络
deformation prediction
extreme learning machine
linear mode solving
elastic network
