中文摘要：

基于小波多尺度理论和最小二乘支持向量机的优越性能，提出了多尺度最小二乘小波支持向量回归，弥补了普通最小二乘小波支持向量回归在单尺度小波空间上对函数进行逼近的不足，使用多尺度上的小波线性组合来逼近L^2（R^d）空间上的任意函数，真正意义上实现了小波分解和最小二乘支持向量机的最佳结合，更有效地继承了小波多尺度学习算法和最小二乘支持向量机的优点，既能达到有效精度而且还计算简便。本文以两尺度为例，通过仿真实验说明了所提算法的有效性。

英文摘要：

Because of the theory of multi--scale wavelet and the superiority of least squares support vector machines, multi--scale least squares wavelet support vector regression is proposed in this paper. Making up for the shortage of the approximation of the function in a single--scale wavelet space by ordinary least squares wavelet support vector regression multi--scale least squares wavelet support vector regression uses the linear combination of the wavelet basis in different space to approximate any function in the space of L^2 （R^d） , realizes the best combination of the wavelet decomposition and least squares support vector machines in the true sense, inherits the advantages of learning algorithm of multi--scale wavelet and least squares support vector machines more effectively, that multi--scale least squares wavelet support vector regression can achieve an effective accuracy and calculate simply simultaneously. In this paper, two--scale least squares support vector regression is performed, and the simulation experiments show the proposed method is effective.