Natural Gas Futures Price Prediction Based on Variational Mode Decomposition–Gated Recurrent Unit/Autoencoder/Multilayer Perceptron–Random Forest Hybrid Model DOI Open Access
Haisheng Yu,

Shenhui Song

Sustainability, Год журнала: 2025, Номер 17(6), С. 2492 - 2492

Опубликована: Март 12, 2025

Forecasting natural gas futures prices can help to promote sustainable global energy development, as the efficient use of a clean source has become key growing demand for development. This study proposes new hybrid model prediction prices. Firstly, original price series is decomposed, and subsequences, along with influencing factors, are used input variables. Secondly, variables grouped based on their correlations output variable, different models employed forecast each group. A gated recurrent unit (GRU) captures long-term dependence, an autoencoder (AE) downscales extracts features, multilayer perceptron (MLP) maps complex relationships. Subsequently, random forest (RF) integrates results obtain final prediction. The experimental show that mean absolute error (MAE) 0.32427, percentage (MAPE) 10.17428%, squared (MSE) 0.46626, root (RMSE) 0.68283, R-squared (R²) 93.10734%, accuracy rate (AR) 89.82572%. demonstrate proposed decomposition–selection–prediction–integration framework reduces errors, enhances stability through multiple experiments, improves efficiency accuracy, provides insights forecasting.

Язык: Английский

Natural Gas Futures Price Prediction Based on Variational Mode Decomposition–Gated Recurrent Unit/Autoencoder/Multilayer Perceptron–Random Forest Hybrid Model DOI Open Access
Haisheng Yu,

Shenhui Song

Sustainability, Год журнала: 2025, Номер 17(6), С. 2492 - 2492

Опубликована: Март 12, 2025

Forecasting natural gas futures prices can help to promote sustainable global energy development, as the efficient use of a clean source has become key growing demand for development. This study proposes new hybrid model prediction prices. Firstly, original price series is decomposed, and subsequences, along with influencing factors, are used input variables. Secondly, variables grouped based on their correlations output variable, different models employed forecast each group. A gated recurrent unit (GRU) captures long-term dependence, an autoencoder (AE) downscales extracts features, multilayer perceptron (MLP) maps complex relationships. Subsequently, random forest (RF) integrates results obtain final prediction. The experimental show that mean absolute error (MAE) 0.32427, percentage (MAPE) 10.17428%, squared (MSE) 0.46626, root (RMSE) 0.68283, R-squared (R²) 93.10734%, accuracy rate (AR) 89.82572%. demonstrate proposed decomposition–selection–prediction–integration framework reduces errors, enhances stability through multiple experiments, improves efficiency accuracy, provides insights forecasting.

Язык: Английский

Процитировано

0