Machine Learning-Based Predictive Analytics for Sustainable Resource Consumption Forecasting: A Comparative Study

Abstract: Accurate prediction of resource consumption has emerged as a fundamental prerequisite for achieving global sustainability targets. Traditional statistical models, while useful, face inherent limitations in capturing the non-linear and multi-dimensional nature of resource usage dynamics. This study investigates the comparative predictive performance of five widely adopted supervised machine learning algorithms—Linear Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), and Gradient Boosting—applied to a structured sustainability dataset comprising population density, industrial activity, energy consumption, water usage, rainfall, and recycling rate variables. Models were trained on an 80:20 stratified data split with standardized feature scaling, and evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the Coefficient of Determination (R²). Empirical results reveal that Linear Regression achieved the highest R² score of 0.9757 with the lowest MAE of 3.7376, followed by Gradient Boosting (R² = 0.9486) and Random Forest (R² = 0.9203). Feature importance analysis confirmed that industrial activity index and energy consumption exert the dominant influence on resource demand predictions. The findings provide data-driven guidelines for policymakers and planners seeking to adopt machine learning-based forecasting frameworks for improved sustainable resource management.

Keywords: Machine Learning; Sustainable Resource Management; Predictive Analytics; Gradient Boosting; Random Forest; Feature Engineering; Comparative Evaluation