基于相空间重构的GA-SVR组合模型边坡位移预测研究

摘要

在传统支持向量回归机的基础上,考虑观测数据的混沌特性,通过对训练样本的相空间重构,并结合遗传算法在寻参上的优势,建立边坡变形的相空间重构GA-SVR组合模型.通过组合模型对某矿山边坡位移预测值与实测值进行对比分析,发现组合模型在预测精度上更具优势.%On the basis of the traditional support vector regression machine,considering the chaotic properties of observation data,the GA-SVR combination model is built by combining the reconstruction of a phase space of training sample and the advantages of a genetic algorithm in seeking the optimum parameter.After comparing and analyzing the predicted and measured values of slope deformation,we determine that the combination model has higher prediction accuracy.