Efficient Gasoline Spot Price Prediction Using Hyperparameter Optimization and Ensemble Machine Learning Approach

Md. Amir Hamja, Md Rakinus Sakib, Mahmudul Hasan and Md Sabir Hossain ()
Additional contact information
Md. Amir Hamja: Hajee Mohammad Danesh Science and Technology University
Md Rakinus Sakib: Hajee Mohammad Danesh Science and Technology University
Mahmudul Hasan: Deakin University
Md Sabir Hossain: King Fahd University of Petroleum and Minerals (KFUPM)

A chapter in Machine Learning Technologies on Energy Economics and Finance, 2025, pp 285-313 from Springer

Abstract: Abstract Energy price prediction is crucial for optimizing resource allocation, reducing costs, and ensuring efficient energy management in markets and industries. To accurately capture the varied fluctuation patterns in energy prices, we introduce an energy price prediction system that integrates Machine Learning (ML), Deep Learning (DL), and Ensemble Learning models. We designed four stacking ensemble ML models to combine the unique features of different ML models, aiming for better prediction performance. Among all the ensemble models, the stacking LRSDR model, (base models: LR, Ridge, SVR, and DT and meta-model: Ridge) which combines Linear Regression (LR), Ridge Regression, Decision Tree Regression (DTR), and Support Vector Regression (SVR), demonstrates superior performance in terms of loss, prediction score, and computational speed. It achieves an R 2 $$^2$$ of 98% for U.S. gasoline prices and 99.2% for New York gasoline prices, with processing times of just 1.57 seconds for U.S. gasoline and 0.66 seconds for New York gasoline. Moreover, low values for MAE (Mean Absolute Error), MSE (Mean Squared Error), RMSE (Root Mean Squared Error), MAPE (Mean Absolute Percentage Error), and sMAPE (Symmetric Mean Absolute Percentage Error) by performing hyperparameter tuning process to find the better parameter values of the ML models to accelerate the training process. To assess the LRSDR model’s effectiveness, we compared it with established individual and ensemble learning methods typically employed for energy price forecasting, such as LR, Ridge, Lasso, Polynomial, SVR, DTR, Gradient Boosting, LightGBM (Light Gradient Boosting Machine), XGBoost (Extreme Gradient Boosting), MLP (Multilayer Perceptron), GRU (Gated Recurrent Unit), LSTM (Long Short-Term Memory), and BiLSTM (Bidirectional Long Short-Term Memory). This study offers critical insights into accurate energy price prediction, essential for optimizing market strategies and enhancing energy efficiency.

Keywords: Energy price forecasting; Hyperparameter optimization; Stacking ensemble learning; Gasoline spot price analysis (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

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:isochp:978-3-031-94862-6_12

Ordering information: This item can be ordered from
http://www.springer.com/9783031948626

DOI: 10.1007/978-3-031-94862-6_12

Access Statistics for this chapter

More chapters in International Series in Operations Research & Management Science from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().