Skip to Content
EntropyEntropy
  • Feature Paper
  • Article
  • Open Access

28 January 2026

18 Pages

Tuning Deep Learning for Predicting Aluminum Prices Under Different Sampling: Bayesian Optimization Versus Random Search

and
Department of Supply Chain and Business Technology Management, John Molson School of Business, Concordia University, Montreal, QC H3H 0A1, Canada
*
Author to whom correspondence should be addressed.
This article belongs to the Section Multidisciplinary Applications

Abstract

This work implements deep learning models to capture non-linear and complex data behavior in aluminum price data. Deep learning models include the long short-term memory (LSTM) and deep feedforward neural networks (FFNN). The support vector regression (SVR) is employed as a base model for comparison. Each predictive model is tuned by using two different optimization methods: Bayesian optimization (BO) and random search (RS). All models are tested on daily, weekly, and monthly data. Three performance metrics are used to evaluate each forecasting model: the root mean squared error (RMSE), mean absolute error (MAE), and the coefficient of determination (R2). The experimental results show that the LSTM-BO is the best-performing model across the time horizons (daily, weekly, and monthly). By consistently achieving the lowest RMSE, MAE, and highest R2, the LSTM-BO outperformed all the other models, including SVR-BO, FFNN-BO, LSTM-RS, SVR-RS, and FFNN-RS. In addition, predictive models utilizing BO regularly outperformed those using RS. In summary, LSTM-BO is highly beneficial for aluminum spot price forecasting.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.