The Combined Forecasting Method of Carbon Trading Price in Pilot Carbon Market Based on the Optimal Weighting Method

In order to effectively improve the prediction accuracy of carbon price and ensure stable and healthy development of the carbon market, in this paper, five pilot carbon markets in China were selected as the example, the Dynamic Time Warping (DTW) method was firstly used to extract critical influence factors of carbon price. Secondly, four typical single prediction models, including ARIMA, LSTM and XGBoost, were established for predicting the carbon price of five pilot carbon markets. Finally, the combined carbon price prediction model was formulated based on optimal weighting method, which innovatively takes predicted results provided by above three single models as the input. The prediction results demonstrated that compared with the single models, the MAPE and RMSE of the combined prediction model are reduced by $0.0002-0.3742$ and $0.0469-14.6960$, respectively, which reflected the superiority and universal applicability of combined prediction model.