Abstract
Proposed research presents a data-driven framework for forecasting municipal solid waste (MSW) generation and emission dynamics in Erode City, India, by employing supervised machine learning algorithms. Leveraging a five-year dataset (2019-2024) comprising socio-economic variables, zonal waste typologies, and historical waste volumes, the model integrates Support Vector Machine (SVM), Random Forest (RF), and Naive Bayes (NB) classifiers. Feature selection and proximity ranking techniques were applied to identify high-impact variables, with plastic and organic waste emerging as dominant predictors. Data pre-processing included normalization, missing value imputation, and spatial zoning analysis. The model was validated through cross-validation with an 80:20 training-to-testing ratio. Among the tested models, SVM exhibited Superior performance, achieving a prediction accuracy of 96%, lowest mean squared error (MSE = 4860), and minimal computational latency (0.67 seconds), indicating suitability for real-time deployment. The integration of proximity matrix analysis and zonal feature clustering enhanced interpretability and robustness. The proposed framework demonstrates significant potential for scalable waste forecasting applications, enabling emission quantification and strategic decision-making. Future work includes the incorporation of real-time sensor data, temporal decomposition, and hybrid deep learning architectures to optimize waste handling and carbon mitigation strategies.