改进的SVR在大坝边坡位移预测中的应用

Application of Improved SVR in Prediction of Side Slope Displacement of the Dam

针对传统支持向量机参数寻优的不足导致大坝边坡位移预测精度低的问题,本文提出了先粗搜,再多次细寻的改进网格参数寻优法。该法建立了SVR大坝边坡位移预测模型,并应用到大坝边坡位移预测。结果表明:改进的SVR大坝边坡位移预测模型的预测精度比传统支持向量机大坝边坡位移预测模型预测的精度高。

Aiming at the problem of low prediction accuracy for side slope displacement of the dam caused by the insufficient of parameters optimization of traditional support vector machine （SVM） , this paper proposed an improved grid parameter optimization method, which required to conduct a coarse search first, followed by several times of fine search.Based on the improved grid parameter optimization method, we established a prediction model for side slope displacement of SVR dam, and applied it in the prediction of side slope displacement.The results showed that the prediction model, established based on the improved method for predicting side slope displacement of SVR dam, had better prediction accuracy than the prediction model basing on traditional SVM.