An ICA-based support vector regression scheme for forecasting crude oil prices

The fluctuations of crude oil prices affect the economic growth of importing and exporting countries as well as regional security and stability. The intrinsic complex features of oil prices and the uncertainty in economic policy pose challenge on the accurate forecasting of crude oil prices. This paper employs independent component analysis (ICA) to analyze crude oil prices which are decomposed into several independent components corresponding to different types of influential factors affecting oil price. We also propose a novel ICA-based support vector regression scheme, namely ICA-SVR2, for forecasting crude oil prices. The ICA-SVR2 starts from the use of ICA to decompose oil price series into three independent components, which are respectively forecasted by SVR models. The forecasted independent components are then integrated together by developing a new SVR model with independent components as inputs for forecasting crude oil prices. Our experimental results show the usefulness of ICA in identifying the driving factors behind the fluctuations of crude oil prices. A comparative study between ICA-SVR2 and other two models shows that ICA-SVR2 is an effective tool in forecasting crude oil prices. (C) 2016 Elsevier Inc All rights reserved.