Exploring Ecological Footprint Drivers in Africa: Fresh Insights from Advanced Machine Learning Techniques

Africa faces escalating ecological challenges driven by rapid urbanization, demographic changes, and the intensifying exploitation of its natural capital. In line with this motivation, analysis of ecological footprint (efp) drivers is revisited using robust and diverse machine learning techniques on a panel dataset of 52 African countries. By implementing in-sample and out-of-sample analyses in predicting key efp drivers, the study evaluates the predictive performance of the outperforming model among six machine learning models—Random Forest (RF), XGBoost, Gradient Boosting Machines (GBM), Support Vector Machines (SVM), Adaptive LASSO, and Histogram-Based Gradient Boosting (HGB). According to the results, the RF model demonstrates the highest predictive power compared to all other models. By incorporating the cumulative importance-based feature selection technique into the RF prediction, our analysis identifies agricultural production, GDP per capita, population density, renewable energy consumption, trade openness, and income inequality as the most influential factors, collectively accounting for 70% of the total variance in the efp across the continent. Additionally, there exists non-linear relationship between ecological footprint and its determinants such as GDP (turning point at ~$6,460.48), agricultural output (with a turning point at 26.22% of GDP), and renewable energy consumption (turning point at 75% of total energy consumption). However, a cross-country heterogeneity analysis using the Phillips and Sul convergence algorithm reveals the complexity of ecological sustainability in Africa, with distinct variables driving ecological footprints across clubs. This finding highlights the limitations of a one-size-fits-all approach to advancing sustainability across the continent.