稀疏最小二乘支持向量机及其应用研究

Sparse Least Squares Support Vector Machine and Its Applications

提出一种构造稀疏化最小二乘支持向量机的方法.该方法首先通过斯密特正交化法对核矩阵进行简约,得到核矩阵的基向量组;再利用核偏最小二乘方法对最小二乘支持向量机进行回归计算,从而使最小二乘向量机具有一定稀疏性.基于稀疏最小二乘向量机建立了非线性动态预测模型,对铜转炉造渣期吹炼时间进行滚动预测.仿真结果表明,基于核偏最小二乘辨识的稀疏最小二乘支持向量机具有计算效率高、预测精度好的特点.

A method is proposed to construct sparse least squares support vector machines （LSSVMs）. Firstly, the kernel matrix is reduced by Schmidt orthogonalization to get base vectors of the kernel matrix. Then, regression of the LSSVM is computed with kernel partial least squares so that the LSSVM is sparse. A nonlinear dynamic prediction model based on sparse LSSVM is constructed to predict the converting time of copper converter at the slag making stage. The simulation results show that the sparse LSSVM based on kernel partial least squares identification is of high computation efficiency and good prediction accuracy.