基于支持向量回归的非线性时间序列预测

Nonlinear Time Series Forecasting Based on Support Vector Regression

针对实际系统的高度非线性及复杂动态性,把非线性时间序列建模与预测问题转换为函数回归估计问题。把具有全局最优性、较好泛化能力及训练效率高的最小二乘支持向量回归算法应用到非线性时间序列预测与建模中。最后给出了某市年电力负荷预测的应用实例,与传统支持向量回归算法相比,文章描述的方法具有较好的预测精度。

The problem of nonlinear time series forecasting is transformed to the problem of function regression according to the nonlinearity and the complexity of actual systems. The Least Squares Support Vector Regression which has advantages such as global optimization and good generalization is used in nonlinear time series forecasting. Finally, an example of power load forecasting of one year is given. The experiments showed Least Squares Support Vector Regression had better forecasting accuracy than Support Vector Regression.