An accurate prediction of crude oil prices over long future horizons ischallenging and of great interest to governments, enterprises, and investors.This paper proposes a revised hybrid model built upon empirical modedecomposition (EMD) based on the feed-forward neural network (FNN) modelingframework incorporating the slope-based method (SBM), which is capable ofcapturing the complex dynamic of crude oil prices. Three commonly usedmulti-step-ahead prediction strategies proposed in the literature, includingiterated strategy, direct strategy, and MIMO (multiple-input multiple-output)strategy, are examined and compared, and practical considerations for theselection of a prediction strategy for multi-step-ahead forecasting relating tocrude oil prices are identified. The weekly data from the WTI (West TexasIntermediate) crude oil spot price are used to compare the performance of thealternative models under the EMD-SBM-FNN modeling framework with selectedcounterparts. The quantitative and comprehensive assessments are performed onthe basis of prediction accuracy and computational cost. The results obtainedin this study indicate that the proposed EMD-SBM-FNN model using the MIMOstrategy is the best in terms of prediction accuracy with accreditedcomputational load.
展开▼
