Carbon Futures Price Forecasting Based On Improved Support Vector Machine Under Multifractal Principle

At present,the international carbon market prices are characterized by nonlinear,non-stationary and multi-frequency irregularities.The lack of timely and effective carbon market price forecast is one of the serious causes of carbon asset loss in China.At the same time,it also makes the carbon market price fluctuation characteristics and The study of forecasting methods is particularly necessary and urgent.Therefore,this paper uses the EUA carbon futures of the European Climate Exchange as the research object.Based on the systematic analysis of the fractal theory and the improvement of the multifractal method,a prediction model based on the multifractal principle is proposed.The main contents and conclusions are as follows:1.Research on Multifractal Features of Carbon Futures Firstly,it summarizes the research on the characteristics and prediction of the volatility behavior of carbon futures price by domestic and foreign scholars and briefly reviews the limitations of the existing literature.Then it discusses the applicability of the fractal market theory to the carbon futures market.(1)This paper firstly overcomes the defects of the MFDFA method(multiple fractal elimination trend fluctuation method)by introducing Empirical Mode Decomposition(EMD).The result confirms that the carbon futures price has long-term memory and multiple analysis features,confirming the carbon futures price.Predictable;(2)The use of EMD is prone to modal aliasing,which affects the resolving power of multifractal signals and needs further optimization.2.Based on Improved Support Vector Machines and Models Based on the forecasting idea of multifractal theory,a combined EEMD-φPSO-SVM prediction model was constructed,and the prediction length was determined based on the fluctuating period of the EEMD reconstruction scores.Finally,other common prediction models were compared.The results showed that EEMD-φPSO-SVM The model has the characteristics of high operating efficiency and higher prediction accuracy,and is more suitable for short-term prediction of carbon futures prices.The academic contribution of this research is to enrich the research system of international carbon futures price forecasting,and provide a scientific and effective pricing reference method for carbon companies to prevent and monitor carbon market risks.