Deep Learning Approaches for Enhanced Predictive Modeling of Carbon Prices

The increasing greenhouse effect has led international organizations and governments to place greater emphasis on environmental protection, particularly carbon emissions. In response to policies such as Peak Carbon and Carbon Neutral, there are strict regulations on carbon emissions from factories. This makes it imperative to accurately predict carbon emissions and carbon price fluctuations. Machine Learning (ML) and Deep Learning (DL), some of the fastest growing Artificial Intelligence techniques, are seen as the key to accurate price prediction through supercomputing. In this paper, we present a new method that combines feature extraction and non-linear modelling to predict carbon prices through a combination of deep learning and machine learning techniques. The method combines K Nearest Neighbours, Completely Integrated Empirical Pattern Decomposition with Adaptive Noise, Sample Entropy, Long Short-Term Memory, Stochastic Gradient Descent, and Random Forest.