Comparative Analysis of Polynomial Regression and Machine Learning Models for CO₂ Emissions Prediction

Accurate modeling of CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions is critical for addressing climate change and informing policy decisions aimed at reducing environmental impacts. This study compares the performance of polynomial regression models (degrees 2–5) and advanced Machine Learning (ML) techniques Random Forest, XGBoost, and Generalized Additive Models (GAM) in forecasting CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions across diverse economies. The aim is to evaluate whether polynomial regression, a traditional statistical approach, can effectively model CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> trends in comparison to more sophisticated ML models that handle non linear relationships and large datasets. The research employs a quantitative experimental approach, using historical CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions, GDP, and population data from seven countries. Model performance is assessed through several metrics, including R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, MSE, RMSE, AIC, and BIC. The results indicate that higher degree polynomial regression models improve accuracy but risk overfitting, particularly in complex datasets. On the other hand, ML models, particularly XGBoost and Gradient Boosting, provide superior predictive performance with lower error rates and greater generalizability. While GAM models offer a flexible and interpretable approach, they do not achieve the same level of accuracy as tree based models. The findings underscore the trade off between interpretability and predictive power in model selection. This study highlights the advantages of ML models for CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions forecasting, suggesting their potential for more reliable and scalable climate policy implementation across various economic contexts.