Carbon Price Prediction With Public Social Media Big Data and an Interpretable Multi‐Objective Intelligent Feature Optimization Strategy

Accurate forecasts of carbon prices are essential for optimizing resource allocation in the carbon market and guiding corporate emissions reduction decisions. However, carbon prices are influenced by a variety of factors, and traditional forecasting methods often fail to account for the complex interrelationships among these factors, making it difficult to extract effective features and reduce the forecasting accuracy. To address these challenges, this study innovatively develops an interpretable intelligent feature optimization strategy based on an improved multi‐objective secretary bird optimization algorithm. This strategy effectively addresses the issue of feature masking by introducing the discrete optimization and bidirectional propagation technique, thereby enabling the precise identification and quantification of useful features. The innovative development of embedded interpretable mechanisms can provide a transparent and interpretable basis for carbon price forecasting in the feature optimization process. Furthermore, this study is the early attempt to incorporate public social media big data, which responds to investor sentiment and attention, into carbon price forecasts, whose unique real‐time and interactive nature can keenly capture social dynamics, further optimizing the timeliness of carbon price prediction. Empirical studies show the proposed feature optimization strategy and the introduction of multimodal public social media big data can significantly improve the precision and robustness of the prediction. The study offers methodological innovations for carbon price forecasting and serves as a valuable reference for investors to optimize their trading decisions.