Beyond Linear Models: Evaluating Tree-Based, Instance-Based, and Deep Learning Methods for Carbon Market Forecasting
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Abstract Transition to the zero-carbon economy is intensely driven by the trade-off between the costs of investing in zero carbon technologies and purchasing emission allowance contracts, since carbon pricing plays a pivotal role in determining the pace and success of this transition. In this context, it is extremely significant to identify substantial carbon price drivers and develop precise carbon price prediction models. Accordingly, this study proposes a comprehensive and comparative research through developing optimized deep learning, tree and instance based models for feature screening of carbon price drivers and predicting the carbon price. Results identify significant impacts of the energy prices and stock returns on the carbon price volatility and determine the coal price, United Kingdom 10-year bond yield, and the Standard and Poor’s 500 index as the leading carbon price drivers. Finally, assessment and validation results highlight the robust precision and reliability features of the Extra Trees (ET), K-nearest Neighbor (KNN), and Deep Neural Network (DNN) methods.
Fields of Study & Subjects
Stock Market Forecasting Methods
Market Dynamics and Volatility
Climate Change Policy and Economics
Publication details
- Journal / source
- Research Square
- Publication date
- 28 March 2026
- Language
- EN
- Document type
- Preprint