基于分解集成及不确定理论的碳价格预测OA北大核心CSTPCD

Accurate carbon market price predictions were the basis for carbon emission trading market-related policy formulation and carbon finance development.In order to eliminate the nonlinearity,non-stationarity,high noise and uncertainty of the original carbon market price series and accurately predict the carbon market price,this paper combined uncertainty theory,ensemble empirical mode decomposition(EEMD)and RBF neural network.Firstly,the original carbon market price data was decomposed and reconstructed by the EEMD algorithm and the fine-to-coarse method,and the high-frequency and low-frequency terms with different changing laws were obtained,and they were substituted into the RBF neural network for training.Then the uncertainty theory was used to analyze the uncertainty of the output weight of the low-frequency term,and the residual trend term was fitted by linear regression.Finally,the forecast results of the three sub-items were integrated and summed to obtain the final carbon market price forecast value.The empirical results showed that whether in terms of RMSE,MAE or MAPE indicators,the model in this paper had more advantages than other prediction models in carbon market price prediction,and the prediction results were more accurate.