基于插值组合核的LS-SVR板形预测方法

基于RBF核的LS-SVR模型,采用实验变差函数计算插值算法权值,对权系数与各训练样本内积值相乘所反映样本空间结构和相互间关联度的插值核函数进行构造,提出了一种通过Kriging空间散乱插值方法利用样本数据构造出的插值核函数与RBF核函数进行组合而成的核方法。结果表明,该方法使LS-SVR板形预测有更好的性能,在提升预测算法泛化能力的同时,实现了对板形的精准回归预测。

Crown prediction for LS-SVR based on interpolated mixture of kernels

Based on LS-SVR with RBF kernel, this paper uses the experimental variogram to calculate the weight of the interpolation algorithm. It constructs the interpolation kernel which reflects the spatial structure and inter-correlation within each training sample weight and inner product value multiplied. And it proposes the mixture of kernels that combines RBF kernel and interpolation kernel constructed with sample data by the Kriging scattered data interpolation method. The results show that the method improves the performance of LS-SVR crown prediction. It not only enhances the generalization ability of the prediction algorithm but also achieves the precise regression of crown prediction.