Multistep-ahead forecasting of coal prices using a hybrid deep learning model. (March 2020)

• Record Type: Journal Article
• Authors: Alameer, Zakaria; Fathalla, Ahmed; Li, Kenli; Ye, Haiwang; Jianhua, Zhang
• Is Part Of: Resources policy. Volume 65(2020)
• Publication Date: 2020-03
• Subjects: Coal price fluctuations -- Deep learning -- DNN -- LSTM -- Multistep-ahead forecasting; Mines and mineral resources -- Periodicals; Ressources minérales -- Périodiques; Ressources naturelles -- Gestion -- Périodiques; Environnement -- Politique gouvernementale -- Périodiques; 333.8
• DOI: 10.1016/j.resourpol.2020.101588
• Languages: English
• ISSNs: 0301-4207

Abstract

Abstract: An accurate forecasting model for the future coal price fluctuations provides critical information and early warning for government and policymakers to provide a stable supply of energy which considered the main concern for China's policymakers. This paper proposes a deep learning model for accurately forecasting monthly coal price fluctuations at different horizons. The proposed LSTM–DNN model combines long-short term memory (LSTM) and deep neural network (DNN). To demonstrate how LSTM–DNN model leads to improvements in predictive accuracy. We compared the results of the LSTM–DNN model to other competitive models, which include multilayer perceptron neural network (MLP) and support vector machine (SVM). Experimental results show the superiority of the hybrid LSTM–DNN model over the other competitive models and its capability to forecast multiple steps accurately and flexible. Therefore, the proposed approach represents an effective and promising technique for the long-term future prediction of coal price fluctuations. Highlights: A new deep learning model to forecast coal price is presented. The model leads to accuracy improvements that are statistically significant. The correlation analysis between coal price and predictor variables is investigated. The proposed approach can be applied in the international energy market.