Predicting Oil Price Movements: A Dynamic Artificial Neural Network Approach

Ali Abbasi Godarzi*, Rohollah Madadi Amiri, Alireza Talaei, Tooraj Jamasb

*Corresponding author for this work

Research output: Contribution to journalJournal articleResearchpeer-review


Price of oil is important for the economies of oil exporting and oil importing countries alike. Therefore, insight into the likely future behaviour and patterns of oil prices can improve economic planning and reduce the impacts of oil market fluctuations. This paper aims to improve the application of Artificial Neural Network (ANN) techniques to prediction of oil price. We develop a dynamic Nonlinear Auto Regressive model with eXogenous input (NARX) as a form of ANN to account for the time factor. We estimate the model using macroeconomic data from OECD countries. In order to compare the results, we develop time series and ANN static models. We then use the output of time series model to develop a NARX model. The NARX model is trained with historical data from 1974 to 2004 and the results are verified with data from 2005 to 2009. The results show that NARX model is more accurate than time series and static ANN models in predicting oil prices in general as well as in predicting the occurrence of oil price shocks.
Original languageEnglish
JournalEnergy Policy
Issue numberMay
Pages (from-to)371-382
Number of pages12
Publication statusPublished - 2014
Externally publishedYes


  • Oil price forecasting
  • Time series model
  • NARX model

Cite this