Emission Trajectories in the Trade Context: A Comprehensive Machine Learning Approach Using K-Means and ARIMA

ARIMA modeling, K-means clustering, and Correlation analysis are applied in this study to predict “emission per capita” and “merchandise trade” as percentage of GDP for 10 developed countries. The objective is to recognise and forecast trends in emissions and trade flows, and to cluster countries with similar emissions, population and trade activity. Countries were categorized into three clusters according to output of K-means clustering based on mean per capita emissions (mean emissions per capita range from 2.1 to 9.8 metric tons per capita) and merchandise trade as percentage of GDP (20% to 55%). An ARIMA model was then built for each cluster to predict future trends in emissions. For Cluster 1 (low-emission countries), emissions reductions are anticipated to fall between 1.5% and 3.2% in the next 10 years; for Cluster 2 (moderate-emission countries), between 2.0% and 4.5%. Cluster 3 (high emissions) will lower emissions by between 0.5 and 2.1%. Results reveal how trends in emissions are likely to change over the next 10 years for groups of countries associated with similar levels of activity. It highlights the performance of clustering methods for prediction of environmental economics - a critical topic with ramifications for decision-making in emissions reduction and trade.