Hybrid data-driven assessment and optimization strategies for municipal solid waste generation

Area of Science:

• Environmental Science
• Data Science
• Urban Planning

Background:

• Per-capita waste generation (PCWG) is a critical metric for urban sustainability.
• Understanding PCWG drivers is essential for effective waste management strategies in India.
• Existing datasets lack harmonization and comprehensive analytical frameworks.

Purpose of the Study:

• To develop a harmonized dataset and an interpretable analytical framework for PCWG assessment across Indian states.
• To identify key factors influencing PCWG using advanced machine learning and explainability techniques.
• To provide a resource for comparative analysis, policy development, and urban sustainability studies.

Main Methods:

• Data harmonization and preprocessing for municipal solid waste data.
• Hybrid feature selection incorporating mutual information, tree-based metrics, and SHAP values.
• Supervised machine learning models (CatBoost, Random Forest) for PCWG prediction.
• Bayesian optimization and NSGA-II for scenario exploration and trade-off analysis.

Main Results:

• High predictive accuracy for PCWG with R² values of 0.952 (CatBoost) and 0.802 (Random Forest).
• Urbanization intensity and population density identified as primary drivers of PCWG.
• Pareto-optimal solutions generated for waste generation, treatment capacity, and cost trade-offs.

Conclusions:

• The study provides a reproducible and transferable framework for PCWG analysis.
• Annotated datasets and model outputs support informed policy decisions for urban sustainability.
• The research offers valuable insights for environmental and regional management contexts.