ABSTRACT
The development of a robust, high-quality, and accurate model for forecasting carbon emissions is a prerequisite for providing insights into environmental policies for achieving the Paris Agreement on climate change. Research comparing the forecasting ability of different machine learning (ML) algorithms by focusing on carbon emissions and more specifically on Australian’s carbon emissions remains scarce. This chapter, therefore, applies different ML techniques, such as a decision tree (DT), a random forest (RF), extreme gradient boosting, and support vector regression (SVR), to model Australia’s carbon emissions. The findings indicated that a DT has a higher coefficient of determination (R2) of 99.71%, followed by an RF with an R2 of 99.14%, extreme gradient boosting (XGBoost) with an R2 of 98.88%, and SVR with an R2 of 97.42%. In terms of accuracy, the tree-based models had the lowest errors. Overall, the DT model produced the most accurate predictions. On the other hand, the kernel-based model, Radial Basis Function (RBF)-SVR, had comparatively higher errors. For computational efficiency, the DT and SVR models were more efficient than XGBoost and RF. Comparatively, the DT model ranked first among the other ML techniques utilized in this study, based on the performance assessment metrics.
