Comparative study on carbon price forecast based on LSTM and ARIMA model

With the improvement of China's carbon emissions trading market, more attention is paid to the prediction of trading price. Based on the average monthly transaction price of Beijing carbon market, this paper constructs ARIMA(Autoregressive Integrated Moving Average) model to make a three-month short-term prediction of carbon price, and evaluates the prediction effect of LSTM(Long Short-term Memory) model. By comparing the predicting accuracy of LSTM and ARIMA model, this paper finds that ARIMA model has higher predicting accuracy than LSTM model in short-term prediction of univariate time series data. Traditional statistical methods should be combined with machine learning, deep learning and other artificial intelligence algorithms to improve the prediction ability in practical application. Meanwhile, in order to prevent the price fluctuation risk of carbon emissions trading, government should strengthen the carbon emission quota regulation, control the drastic price fluctuation, integrate regional and national markets and improve the liquidity of carbon emissions trading markets.