Forecasting oil price in times of crisis: a new evidence from machine learning versus deep learning models

Haithem Awijen (), Hachmi Ben Ameur (), Zied Ftiti () and Waël Louhichi ()
Additional contact information
Haithem Awijen: Inseec Grande École, Omnes Education Group
Hachmi Ben Ameur: Inseec Grande École, Omnes Education Group
Zied Ftiti: OCRE Research Laboratory
Waël Louhichi: ESSCA School of Management

Annals of Operations Research, 2025, vol. 345, issue 2, No 17, 979-1002

Abstract: Abstract This study investigates oil price forecasting during a time of crisis, from December 2007 to December 2021. As the oil market has experienced various shocks (exogenous versus endogenous), modelling and forecasting its prices dynamics become more complex based on conventional (econometric and structural) models. A new strand of literature has been attracting more attention during the last decades dealing with artificial intelligence methods. However, this literature is unanimous regarding the performance accuracy between machine learning and deep learning methods. We aim in this study to contribute to this literature by investigating the oil price forecasting based on these two approaches. Based on the stylized facts of oil prices dynamics, we select the support vector machine and long short-term memory approach, as two main models of Machine Learning and deep learning methods, respectively. Our findings support the superiority of the Deep Learning method compared to the Machine Learning approach. Interestingly, our results show that the Deep LSTM-prediction has a close pattern to the observed oil prices, demonstrating robust fitting accuracy at mid-to-long forecast horizons during crisis events. However, our results show that SVM machine learning has poor memory ability to establish a clearer understanding of time-dependent volatility and the dynamic co-movements between actual and predicted data. Moreover, our results show that the power of SVM to learn for long-term predictions is reduced, which potentially lead to distortions of forecasting performance.

Keywords: Oil; Support vector machines; Long short-term memory; Forecast; Crisis; C14; P28; G12; G13 (search for similar items in EconPapers)
Date: 2025
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10479-023-05400-8 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:annopr:v:345:y:2025:i:2:d:10.1007_s10479-023-05400-8

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10479

DOI: 10.1007/s10479-023-05400-8

Access Statistics for this article

Annals of Operations Research is currently edited by Endre Boros

More articles in Annals of Operations Research from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().