A multi-factor integrated model for carbon price forecasting: Market interaction promoting carbon emission reduction

Abstract

Reasonable carbon price can effectively promote the low-carbon transformation of economy. The future carbon price has an important guiding significance for enterprises and the country. However, the nonlinear and high noise characteristics inherent in carbon price make them challenging to predict accurately. A hybrid decomposition and integration prediction model is proposed using the Hodrick-Prescott filter, an improved grey model and an extreme learning machine to solve this problem. First, a large number of factors that influence carbon price are collected by meta-analysis. The final input is selected through a two-stage feature selection process. Second, the HP filter is used to decompose the input into long-term trends and short-term fluctuations predicted by the improved GM and ELM, respectively. Finally, the two prediction sequences are compared to obtain the final result. European Union Allowances futures price data are applied for empirical analysis. The results show that the prediction performance of this model is better than the other 10 benchmark models, the T-bill, Stoxx50, S&P clean energy index and Brent oil price in the financial and energy markets are helpful in the carbon price's prediction. T-bill affects carbon price frequently, Stoxx50 has a negative correlation with the carbon price in the influence period. Under normal circumstances, the S&P clean energy index is positively correlated with the carbon price. However, when the economic situation is depressed, resulting in a short-term negative correlation between them. In general, carbon market is significantly affected by cross spill over between different markets. The method not only improves the accuracy of carbon price forecast, but also the application of the improved GM explains the reasons for the change of carbon price, which is helpful to promote the realization of carbon neutralization by market-oriented means.