Unstructured Data-driven Hybrid Decomposition-ensemble Carbon Price Combined Forecasting Approach And Empirical Research

In recent years,air pollution,climate change,melting glaciers,and the greenhouse effect have become the focus of government,experts,scholars,and the general public.One of the main causes of these environmental problems is excessive greenhouse gas emissions.Among them,accurate prediction of the carbon price plays a vital role in reducing greenhouse gas emissions.However,the carbon price has high non-linear and non-stationary characteristics,which makes carbon price prediction face huge challenges.The combination forecasting model based on decomposition method can transform time series into stable and regular sub-sequences,so as to better grasp the carbon price fluctuation characteristics.In addition,with the rapid development of the Internet,more and more people search and make information through the Internet,and constantly network data for carbon price forecasting provides a new perspective for research.This paper proposes an unstructured data-driven hybrid decomposition-ensemble carbon price combined forecasting approach.First,select historical data of the carbon price,determine the influencing factors of structured data from the supply and demand of carbon prices,and extract keywords from Baidu Index as unstructured data.Second,inputting all the keywords based on Web search as carbon price input will lead to problems such as information redundancy and heavy calculations.Therefore,the data is reduced by principal components.Then,the historical data,structured data and unstructured data are decomposed and reconstructed by hybrid decomposition method.This process can decompose carbon prices into stable subsequences and contain more effective information.At this time,statistical method and intelligent algorithm can better fit the trend of carbon prices,and obtain carbon prices forecasting results based on unstructured data.Finally,in order to verify the validity and applicability of the proposed model,the carbon price of Hubei Province is predicted,and various error indicators are used for evaluation.This paper first combines a hybrid decomposition and ensemble method with unstructured data to forecast carbon price frequency sequences.Through empirical analysis,it is demonstrate that this model can significantly improve the forecasting accuracy of carbon prices.