Assessing the Relative Importance of the Drivers of CO2 Emissions in the Selected Emerging Economies Using Machine Learning Approach

The main objective of the present research is to answer a key research question: what is the relative importance of the drivers of CO2 emissions? Another important question the present study addresses is how the countries are related to each other regarding CO2 emissions. Taking a sample of 42 emerging economies from Asia and Sub-Saharan Africa (SSA) and using hierarchical clustering and the neural network method the study tries to answer the key question. Firstly, the explanatory variables were identified through a review of the literature. Subsequently, the gathered data was classified into two clusters having similar characteristic variables utilizing the dendrogram by performing an exploratory clustering method known as hierarchical clustering. Later using the machine learning K-Means clustering technique, the clusters were verified. The use of another machine learning method of feed-forward multilayer perceptron commonly known as neural network helped us to identify the relative importance of explanatory variables viz. economic growth (EG), renewable energy consumption (REC), urbanization (URB), democracy index (DI) and foreign direct investment (FDI) for their relation to the response variable viz. CO2 emissions. The neural network results reveal that EG, REC, and URB are the most important variables (with Rank 1,2, and 3 respectively) followed by DI (Rank 4) and FDI (Rank 5). FDI seemingly is the least important among these identified variables.