LS-SVR with Tuned Hyperparameters in Dam Crack Forecasting

摘要：

This paper deals with the application of least squares support vector regression (LS-SVR) with radial basis function (RBF) kernel in dam crack forecasting. In the process of LS-SVR, we performed the standard grid search and particle swarm optimization (PSO) to tune hyperparameters of LS-SVR. The results demonstrate that onr PSO approach can identify optimal or near optimal parameters faster than the exhaustive grid search. Comparison with results from stepwise regression was also included, to evaluate the reliability of applying such a PSO method which avoids doing an exhaustive grid search. We found that our LS-SVR approach is promising in dam crack forecasting, however it cannot be used to extract the crack contributed by water pressure, temperature variation, and aging effect, respectively.

关键词： LSSVR RBF hyperparameter PSO grid search stepwise

作者: Xu Chang Deng Chengfa

作者单位: Department of Hydraulic Engineering Zhejiang Water Conservancy and Hydropower College Hangzhou P.R.C Zhejiang Guangchuan Engineering Consulting Co.Ltd.Zhejiang lnstiute of Hydraulic & Estuary Hangzhou,

会议类型: 国际会议

会议名称: The 2nd IEEE International Conference on Advanced Computer Control(第二届先进计算机控制国际会议 ICACC 2010)

会议地点: 沈阳

会议语种:英文

页码: 65-69

在线出版日期: 2010-03-27（万方平台首次上网日期，不代表论文的发表时间）