Environmental sustainability performance assessment in relation to visibility in African regions with interpretable machine learning

Abstract

Most African countries have developed their economies without causing large-scale ecological damage, yet they face severe challenges from global climate change and COVID-19 epidemic. To address this paradox, explainable machine learning coupled with panel data models are proposed to examine the different performances of Cameroon, Zimbabwe, Ghana, and Eritrea in terms of environmental sustainability measured by visibility. The results show that all countries except Zimbabwe have inverted N-shaped Environmental Kuznets Curves (EKCs), suggesting that the economic development works against environmental sustainability. Based on the modified TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution), Cameroon’s environmental sustainability performance is particularly poor, while Eritrea’s is better. Meanwhile, Eritrea is not exogenously threatened by epidemics, while Cameroon and Ghana are. By integrating the multidimensional Sustainable Development Goal (SDG) achievement pathway assessment system, Zimbabwe may be moving towards green growth. More importantly, the Shapley additive explanation (SHAP) technique is applied to quantify and visualize the socio-economic drivers of environmental sustainability. The results depict that, in addition to meteorological drivers, per capita income plays a leading role. In conclusion, we can validate the contribution of artificial intelligence algorithms to the SDGs and employ a comprehensive method for assessing SDG performance linking environment, economy, and public health. This could lead to some key recommendations for policymakers and international investors to better contribute to the SDGs.