An interpretable analytical framework for carbon price forecasting: Combining multi-source factors and price decomposition

Effective carbon price analysis is crucial for understanding the carbon pricing mechanism, optimizing resource allocation, and promoting the carbon market's sustainable development. Nevertheless, existing analytical models for carbon price fluctuations often struggle to balance accuracy and interpretability, undermining user trust in decision-making models and limiting their ability to provide sufficient support for policymakers and market participants. To address this challenge, we develop a novel carbon price analysis framework based on an interpretable artificial intelligence integrated method to enhance decision-making quality. This framework achieves accurate carbon price prediction and investment timing through feature selection and intelligent optimization algorithms. Meanwhile, we employ multiple interpretability methods to analyze the marginal contributions of influencing factors to carbon prices. Empirical research on different carbon markets validates the predictive advantages of the proposed integrated framework. Furthermore, through the interpretive analysis of long-term trends and short-term fluctuations in carbon prices across different markets, we find that the factors influencing carbon prices exhibit significant spatiotemporal heterogeneity, interdependence, and lag effects. These results demonstrate that the framework can provide reliable insights into the pricing mechanism and investment strategies for market decision-makers and participants, thereby offering important support for achieving multiple United Nations Sustainable Development Goals.