支持向量机的核函数一直是影响其学习效果的重要因素.本文基于小波分解理论和支持向量机核函数的条件,提出一种多维允许支持向量尺度核函数.该核函数不仅具有平移正交性,且可以以其正交性逼近二次可积空间上的任意曲线,从而提升支持向量机的泛化性能.在尺度函数作为支持向量核函数的基础之上,提出基于尺度核函数的最小二乘支持向量机(LS-SSVM).实验结果表明,LS-SSVM 在同等条件下比传统支持向量机的学习精度更高,因而更适用于复杂函数的学习问题. The kernel function of SVM has always been an important factor affecting its learning effect.Based on the wavelet decomposition theory and the condition of SVM kernel function, a multi-dimensional support vector-scale kernel function is proposed.The kernel function not only has translational orthogonality , And can be approximated by its orthogonality to any curve on the quadratic integrable space to enhance the generalization performance of SVM.On the basis of scale function as kernel of support vector, (LS-SSVM) .Experimental results show that LS-SSVM has higher learning accuracy than traditional support vector machines under the same conditions, which makes it more suitable for learning complex functions.