Scholarlink

Accurate oil price forecasting is crucial for stakeholders in the energy sector, especially during periods of high volatility. This study compares the performance of three deep learning models—Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Transformer-based BERT—on oil price prediction using historical data from 2008 to 2024. Our results show that the CNN model outperforms the others with the lowest RMSE (72.48), MAE (75.16), and MAPE (72.48), demonstrating its ability to capture temporal patterns. The LSTM model follows with an RMSE of 80.85, while the Transformer yields a slightly higher RMSE of 79.97 but performs well during volatile periods. CNN also achieves the highest R2 (96.25%), indicating its strong explanatory power. These findings highlight the potential of CNN for effective oil price forecasting.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparing LSTM, CNN, and BERT Models in Oil Price Forecasting

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hicham Boussatta,
  • Marouane Chihab,
  • Mohamed Chiny,
  • Younes Chihab

摘要

Accurate oil price forecasting is crucial for stakeholders in the energy sector, especially during periods of high volatility. This study compares the performance of three deep learning models—Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Transformer-based BERT—on oil price prediction using historical data from 2008 to 2024. Our results show that the CNN model outperforms the others with the lowest RMSE (72.48), MAE (75.16), and MAPE (72.48), demonstrating its ability to capture temporal patterns. The LSTM model follows with an RMSE of 80.85, while the Transformer yields a slightly higher RMSE of 79.97 but performs well during volatile periods. CNN also achieves the highest R2 (96.25%), indicating its strong explanatory power. These findings highlight the potential of CNN for effective oil price forecasting.