The Forecast Of Carbon Futures’ Prices Based On ARIMA-LSSVM Model

Since the Kyoto Protocol has became effective, many carbon exchanges have been established because of the considerable attention given to carbon emissions. The emergence of these exchanges makes carbon emissions become a tradable commodity and give the price of carbon emissions. The carbon finance which takes carbon emissions as the subject, becomes an important part of the global financial markets. As a new carbon financial derivative, carbon future has been greatly developed. The forecast of carbon future’s price is helpful for improving the bargaining power of carbon emissions and the passive situation in which the carbon emissions trade. By forming a reasonable price of carbon emissions, it can give Chinese enterprises more profits in the reduction of carbon emissions’ trading projects.In this paper, the variance ratio test is used to determine whether the carbon futures market is weak and effective. The prices of carbon futures can be studied only when the carbon futures’ market is effective,and forecast for the prices of carbon futures is more convincing and has a theoretical basis. To the price of carbon futures data, using DFA MF analysis to determine the memory of carbon futures prices, and fractal features. We can know about the nonlinear characteristics of carbon futures’ prices. It’s a big issue which the efficient market hypothesis can not study and solve. It also provides a theoretical support to the selected data’s length and maturity for carbon futures’ price forecast. It can help determine the time span of carbon futures’ price forecast by studying the impact of heterogeneous environment on the price of carbon future. The ARIMA, LSSVM and the hybrid model of ARIMA and LSSVM are used to study the prediction of carbon futures’ prices, and the prediction results are compared according to the statistic of root mean square error(RMSE) and direction prediction(). It overcomes the limitation that we forecast only from one aspect while carbon futures’ prices may contain linear and nonlinear forms. Finally, the prediction of the ARIMA-LSSVM2 model is the best. statD...