基于层级支持向量机的石油价格预测

OIL PRICE FORCASTING BASED ON HIERARCHICAL SUPPORT VECTOR MACHINE

本文针对石油价格预测中数据选择不当、数据没有预处理和预测方法单一等问题,提出了层级支持向量机模型(SVM),予以解决。模型的第一层通过Gauss径向基核的支持向量回归机(SVR)对输入数据进行了预处理;模型的第二层对模型第一层所确定的输入数据进行融合,并做出最终的预测,最后用油价波动趋势进行了拟合分析。实例研究表明,层级SVM方法比SVR模型和BP神经网络的性能指标更优,具有很好的应用前景。

This paper presents a hierarchical support vector machine model(SVM)for oil price forecasting due to inappropriate data,no data preprocessing and single forecasting method, which can solve the problems.First, the input data are processed by support vector regression (SVR )with Gauss radial basis kernel using the first level of model. Then the certain input data are integrated by the second level and the final forecast is gained. At last fitting analysis is carried out by the fluctuating trend of oil price.Case studies show that the performance index of hierarchical SVM method, which has a valuable prospect, is better than the SVR model and BP neural network.