An optimal weight heterogeneous integrated carbon price prediction model based on temporal information extraction and specific comprehensive feature selection

Accurate prediction of carbon price provides a strong guarantee for the prosperity of carbon market and the sustainable development of society. This paper proposes an optimal weight heterogeneous integrated carbon price prediction model based on temporal information extraction and specific comprehensive feature selection. Specifically, improved variational mode decomposition is first employed to segment the original carbon price data into distinct sequences, each capturing unique time scale characteristics. Secondly, by comprehensively considering 12 feature screening methods, specifically obtain the optimal impact factors of each subsequence, achieving the purpose of reducing noise interference and improving input quality. Finally, heterogeneous modeling and optimal weight integration, coupled with the northern goshawk optimization, fully leverage the distinct strengths of each model and achieve advantageous complementarity, thereby bolstering the prediction's robustness and precision. The research results demonstrate that, in comparison to the baseline model, the proposed model has achieved remarkable accuracy enhancements in two markets. Taking the Shanghai market as a case, it achieved mean squared error, mean absolute error and mean absolute percentage error values of 1.5199, 0.846, and 0.0145, respectively, with the highest accuracy improvements reaching 43.5555 %, 48.7471 %, and 49.6552 %, respectively. This underscores the outstanding carbon price prediction capability of the model.