Modeling Carbon Emission by Composite Model of Neural Network and Auto-Regression Integrated Moving Average

In recent decade, increasing CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emission further aggravated greenhouse effect. Most of the models regarding the analysis and prediction of carbon dioxide (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emission were based on the emission intensity, population, economy and technology (IPAT) models and modified IPAT models. By applying multiple linear regression analysis of the three or four parameters to forecast the trend of the emission change in the future years, these models are easy to lead to overfitting and insufficient generalization performance. In this study, we establishes a back propagation neural network (BPNN) model using the parameters: population, GDP, energy consumption and Green Finance Index (GFI) as input variables, and CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emission as output variable. The composite model of Autoregressive integrated moving average (ARIMA) model and BPNN model are combined to forecast the trend of the four input variable and CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions output variable change of Shandong province of China in 2020–2030.