基于改进支持向量机的TVARMA模型辨识

A TVARMA Model Identification Method Based on Modified LS-SVM

提出了一种改进的最小二乘支持向量机并将之应用于时变自回归滑动平均模型的辨识。与传统的最小二乘支持向量机相比,通过同时引入结构风险矩阵Q和经验风险权重因子vi,既降低了数据存储空间,又兼具较好的灵活性和适应性,并成功地应用于TVARMA模型的参数辨识过程。实验结果表明方法的有效性。

An improved Least Squares Support Vector Machine（LS-SVM）was proposed and applied to identify the Time-Varying Auto-Regressive Moving-Average（TVARMA）model.Compared with traditional LS-SVM,the Structural Risk Matrix Q and the Empirical Risk weights vi were combined,which reduced the data space and had satisfying flexibility and adaptability.The method was successfully applied to identify the TVARMA model parameters.The experimental results verified its feasibility.