Forecasting sustainable development indicators using XGBoost: Evidence from Brazil, Canada, China and India (2005–2023)

Abstract

Purpose: This paper discusses the potential of machine learning to predict sustainable development indicators, gross domestic product (GDP), employment, and CO₂ emissions in Brazil, Canada, China and India to build a sustainable policy. Design/Methodology/Approach: An annual panel data (2005–2023) was employed to adopt a deductive and explanatory design. XGBoost (Extreme Gradient Boosting) is an algorithm that models nonlinear relationships and identifies important predictors based on macroeconomic, environmental, and policy variables. Findings: XGBoost revealed high accuracy on GDP, and the balance of payments, climate policy, and green investment were some of the important predictors. CO₂ and employment forecasts were less certain because they were overfitted. Originality/Value: The paper identifies the application of machine learning to predict sustainable development, particularly economic modelling. Research Limitations/Implications: There is an implication of overfitting and data limitations, which implies that higher-frequency data and hybrid models are required. Practical Implications: The implication is based on evidence based policymaking in green investment planning and climate policy evaluation.