Carbon price prediction using multiple hybrid machine learning models optimized by genetic algorithm

Abstract

Importance of the carbon trading has been escalating expeditiously not only because of the environmentalist purposes to mitigate the adverse effects of climate change but also the increasing diversification benefits of the carbon emission contracts due to the low correlation between the emission, equity, and commodity markets. In accordance with the promptly rising significance of accurate carbon price prediction, this paper develops and compares 48 hybrid machine learning models by using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Variational Mode Decomposition (VMD), Permutation Entropy (PE), and multiple types of Machine Learning (ML) models optimized by Genetic Algorithm (GA). The outcomes of this study present the performances of the implemented models at different levels of mode decomposition and the impact of genetic algorithm optimization by comparing the key performance indicators that the CEEMDAN-VMD-BPNN-GA optimized double decomposition hybrid model outperforms the others with a striking R2 value of 0.993, RMSE of 0.0103, MAE of 0.0097, and MAPE of 1.61%.