Monthly crude oil spot price forecasting using variational mode decomposition

Abstract

Crude oil is one of the most important trade commodities in the world and its price fluctuation has significant effects on global economic activities. In this paper, we proposed hybrid models for monthly crude oil price forecasting using variational mode decomposition and artificial intelligence (AI) techniques (in this paper, support vector machine optimized by genetic algorithm (GASVM) and back propagation neural network optimized by genetic algorithm (GABP) are employed for analyzing). In addition, influencing factors of the long-term crude oil price such as the global crude oil production as well as economic activity (Dow Jones Industrial Index is considered in this paper) are investigated and considered on the crude oil price forecasting. Empirical forecasting results of monthly West Texas Intermediate (WTI) and Brent crude oil spot prices validate that the hybrid VMD-based models are superior to previously popular EEMD-based models and single models in terms of both level and directional prediction accuracies as well as the DM test results. In addition, the VMD-AI based models which consider influencing factors of the long-term crude oil price variation perform better than that without considering influencing factors of the long-term crude oil price variation in terms of MAPE and RMSE. All of which confirm that the newly proposed VMD-AI based models are promising tools for crude oil price analysis and forecasting.