Research On Price Forecasting And Trading Strategy Of Futures In Carbon Trading Market

With the intensification of global environmental change, the world common control carbon dioxide emissions. The development of low carbon economy has been valued by many countries, and the corresponding carbon emissions trading market has become a research hotspot. The EU allowances(EUA) futures and certified reduction emission reductions(CER) futures are the main subject matter of the carbon emissions trading market, and are the tools for project owners to avoid risks and obtain profits. Wavelet analysis, genetic algorithm and neural network are used to study the price trends and trading strategies of the EU allowances(EUA) futures and certified reduction emission reductions(CER) futures.The EU allowances(EUA) futures and certified reduction emission reductions(CER) futures related parameters was selected as the research object of carbon futures price prediction and trading strategy in the carbon trading market. The time range of data selected to January 2, 2013 to December 15, 2014. Cointegration test was used to determine whether they have long-term cointegration relationship. The Granger causality test was used to determine the lead lag relationship. Improved wavelet neural network model using genetic algorithm was established to forecast CER futures price trend which had the lag relationship. The experimental results show that the EUA futures and CER futures exist long-term equilibrium cointegration relationship between the futures price, and EUA futures prices ahead of the CER futures prices. Predicted results was measured by four index like the absolute error and the mean absolute percentage error and show that the improved model has higher prediction accuracy that of BP wavelet neural network. Mexican hat wavelet neural network models which using genetic algorithms improved has the best frequency localization characteristic and the best prediction effect, which almost in line with the actual price fluctuations. And the effect of Morlet wavelet model using genetic algorithm improved is a bit poor.The EUA futures prices decomposed and reconstructed to the trend part and detail part by DB6 wavelet. Then, they were learned by genetic-wavelet neural network model which used to delimit the adaptive transaction threshold. Trading strategies was selected by the comparison of daily price and adaptive transaction threshold. Finally, performance of adaptive trading rules in different market was determined by return rate according to the adaptive trading rules under different EUA futures market. Adaptive trading strategy has outstanding performance when EUA futures market volatility and also can get good returns when market volatility is more gentle.The improved wavelet neural network model established in this paper provide an effective way to solve the prediction problem and adaptive trading strategies proposed had good performance both in frequent volatility EUA futures market and relatively stable EUA futures market. This method provides certain reference function for transaction decision-making formulation in the carbon emissions futures market, also provides a new angle of view and the research foundation for research of other futures markets.