Data on forecasting energy prices using machine learning

Gabriel Paes Herrera; Michel Constantino; Benjamin Miranda Tabak; Hemerson Pistori; Jen-Je Su; Athula Naranpanawa

doi:10.1016/j.dib.2019.104122

Data on forecasting energy prices using machine learning

Data Brief. 2019 Jun 12:25:104122. doi: 10.1016/j.dib.2019.104122. eCollection 2019 Aug.

Authors

Gabriel Paes Herrera^{1

2}, Michel Constantino², Benjamin Miranda Tabak³, Hemerson Pistori^{2

4}, Jen-Je Su¹, Athula Naranpanawa¹

Affiliations

¹ Department of Accounting, Finance and Economics, Griffith University, Nathan Campus, Queensland 4111, Australia.
² Department of Environmental Sciences and Sustainability, Dom Bosco Catholic University, Campo Grande, MS, Brazil.
³ School of Public Policy and Government, Getulio Vargas Foundation (EPPG/FGV), Brasilia, DF, Brazil.
⁴ Department of Computer Science, Federal University of Mato Grosso do Sul, Campo Grande, MS, Brazil.

Abstract

This article contains the data related to the research article "Long-term forecast of energy commodities price using machine learning" (Herrera et al., 2019). The datasets contain monthly prices of six main energy commodities covering a large period of nearly four decades. Four methods are applied, i.e. a hybridization of traditional econometric models, artificial neural networks, random forests, and the no-change method. Data is divided into 80-20% ratio for training and test respectively and RMSE, MAPE, and M-DM test used for performance evaluation. Other methods can be applied to the dataset and used as a benchmark.

Keywords: ANN; Coal; Natural gas; Oil.