Title : ( A Hybrid Decomposition and Deep Learning Approach for Forecasting Brent Oil Prices )
Authors: Amirashkan Jafarizadeh , Omid Solaymani Fard ,
Full TextAbstract
This study examines the DLinear model, an advanced deep learning method, for predicting Brent crude oil prices using time series analysis. The data, sourced from the U.S. Energy Information Administration (EIA), covers daily Brent oil Spot Price FOB from May 1987 to March 2025. The model was trained over 836 iterations with a layered early stopping approach, achieving a final loss of 0.0213. Performance metrics include Mean Squared Error (MSE: 0.0219), Mean Absolute Error (MAE: 0.1056), and Root Standard Error (RSE: 0.2679), with a correlation coefficient between predicted and actual values ranging from 0.217 to 0.218. The results highlight the DLinear model’s strong ability to provide accurate forecasts, offering valuable insights for economic planning and energy market analysis. This is due to its skill in handling complex time series through decomposition techniques.
Keywords
, Time series, Forecasting, Brent oil prices, DLinear model, Deep learning, Data analysis
@inproceedings{paperid:1106393,
author = {Jafarizadeh, Amirashkan and Solaymani Fard, Omid},
title = {A Hybrid Decomposition and Deep Learning Approach for Forecasting Brent Oil Prices},
booktitle = {پنجاه و ششمین کنفرانس ریاضی ایران},
year = {2025},
location = {رفسنجان, IRAN},
keywords = {Time series; Forecasting; Brent oil prices; DLinear model; Deep learning; Data analysis},
}
%0 Conference Proceedings
%T A Hybrid Decomposition and Deep Learning Approach for Forecasting Brent Oil Prices
%A Jafarizadeh, Amirashkan
%A Solaymani Fard, Omid
%J پنجاه و ششمین کنفرانس ریاضی ایران
%D 2025
