基于MGM(1,n)模型的多元核支持向量回归预测

Multi-kernel Support Vector Regression Base on MGM(1,n) Model for Time Series Prediction

根据灰色系统和支持向量机相结合的方法,采用多变量灰色模型MGM(1,n)对相互影响、相互制约的多变量时间序列进行模拟,获取残差序列后运用多元核支持向量回归机(MSVR)对残差进行回归以修正原模型,得到多变量灰色支持向量回归复合模型(MGM-MSVR).实证结论表明:复合模型具有比原模型更高的精度.

According to the method of the support vector machine combined with grey system, multi-variable grey model （MGM（1, n）） is adopted to predict multivariate time series, which interact and mutual restrict. After obtaining the residual error sequence, we can carry out multi-kernel support vector regression （MSVR） for residual errors and revise the values of MGM（1, n）. So multi-variable grey composite support vector regression model （MGM - MSVR） is obtained. Experimental results show that composite model has much higher accuracy than the original one.