Global Analysis and Prediction of CO2 and Greenhouse Gas Emissions across Continents

Abstract

Understanding the concentrations of Carbon Dioxide (CO2) and greenhouse gases is very important in solving the problem of climate change. These emissions are the major cause of global warming, which, in turn, has many effects on the environment, economy and society. For this reason, the prediction models for these emissions must be precise to aid policy makers in planning for the effects of the climate in the future. To evaluate the emission data of different continents, this paper seeks to identify related patterns and findings that can help reduce emissions worldwide. The dataset used contains emission data and geographic information from several countries and allows the comparison of several ML models. The models that have been reviewed in this study are linear regression (LR), decision tree regression (DT), random forest regression (RF), support vector regression (SVR), k-nearest neighbor regression (KNN), the XGB regressor, the gradient boosting regressor, Ridge and Lasso. Among the models, the gradient boosting regressor was found to have the best prediction capability, with an R-squared value of 0. The highest value of the mean absolute error (MAE) was 929, and the lowest mean squared error (MSE) was 2535.30. This model outperforms the other models because of its excellent ability to identify the complex interactions between the input variables and emissions. The conclusions stress the possibility of using ensembles, such as gradient boosting, for emission forecasting and present a contribution to studies of this issue for researchers and policymakers. This is a nominal attempt in the ongoing global endeavour to gain insight and curb the determinable levels of CO2 and greenhouse gas emissions for effective decision-maki