基于LS-SVM的基坑周边地表沉降预测模型

针对神经网络用于基坑变形预测存在结构难确定、训练易陷入局部最优及易过学习等问题，以已有的周边地表沉降为样本，利用最小二乘支持向量机（LS—SVM）建立基坑地表沉降预测模型，应用网格搜索算法优化模型参数，时基坑周边地表沉降进行连续滚动的多步预测。实例分析结果表明，LS—SVM用于基坑周边地表沉降预测效果较好，具有所需数据少、推广能力强等优点。

Prediction Model of Ground Settlement of Foundation Ditch Surrounding Based on LS-SVM

The problems that the structure is difficult to indentify, the training is easy to get in local optimization and overfitting are existing in using neural networks to forecast the deformation of foundation ditch. Given this, taking the ground settlement of foundation ditch surrounding as the sample, the prediction model of ground settlement of foundation ditch is established based on LS-SVM, the parameters of the model is optimized by mesh search algorithm, and continuous rolling multi-step forecasting of greund settlement of foundation ditch surrounding is implemented. The result of case analysis indicates that the effect of using LS-SVM to forecast the ground settlement of foundation ditch surrounding is preferable, meanwhile it has such advantages as requiring few data, the capacity of popularization is great.