CO ₂ Emissions Prediction Using Machine Learning Models

Abstract

The effects and repercussions of global warming are profound for people, communities, and the environment. Several activities cause global warming, but carbon dioxide (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions are the main culprit. Humans burn fossil fuels like coal, oil, and gas to produce energy, which releases much carbon dioxide into the atmosphere. To balance (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions in the environment, companies aim toward net zero. Therefore, this work uses a machine learning model to determine the best forecast model for(CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions using the (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions dataset from 1991 to 2020. Machine learning techniques are an effective way to examine the prediction of (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions and have attracted a lot of attention from researchers. The dataset has been divided into an 80:20 train-test (estimation-validation) set and a 20% test set. With varying parameters, Linear Regression, Ridge Regression, LASSO Regression, SVR, and Random Forest Regressor algorithms were used to create the prediction model. The performance of the prediction model was assessed using the error measurement metrics of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Bias Error (MBE), Mean Bias Error (MBE), Coefficient of Determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>), Error squared by the relative root mean (RRMSE), Nash-Sutcliffe Efficiency (NSE), Willmott Index of Agreement (WI) and Root Mean Square Error (RMSE). The best model among the others is the Gradient Boosting Regressor, which yields 0.0049 Root Mean Square Error (RMSE) and 0.0025 Mean Absolute Error (MAE) from the train set.