A novel cluster based multi-index nonlinear ensemble framework for carbon price forecasting.

Forecasting carbon prices has important implications for the formulation of environmental and energy policies as well as for the investment and management of workers in related industries. This paper proposes a novel multi-index nonlinear ensemble framework based on carbon price trading data and technical index data, which combines extreme gradient boost (XGBoost), K-means clustering algorithm based on elbow method (KCE), bidirectional long and short-term memory (BiLSTM) and bidirectional gated recurrent units (BiGRU). First, the XGBoost method is used to filter the raw data set to determine the input variables, and they were then divided into groups using the KCE method. Then, the BiLSTM method is used as the predictor for forecasting the sub model. Finally, BiGRU is applied to nonlinear integration. The developed prediction framework is used to predict China's regional carbon price to evaluate its performance. Experimental results show that the prediction performance of the framework in all target carbon markets is better than the benchmark model, and it is an effective tool for carbon market. [ABSTRACT FROM AUTHOR]