Data-driven artificial intelligence modelling of municipal solid waste generation

Municipal solid waste (MSW) management is a growing challenge for environmental sustainability and public health in urban regions. Reliable forecasts of MSW generation are essential for planning collection systems, treatment capacity and long-term infrastructure, yet conventional approaches often struggle with nonlinear urban dynamics and changing settlement patterns. This study develops an ensemble modelling framework in which Extreme Gradient Boosting (XGBoost) is optimised using six recent metaheuristic algorithms to forecast annual MSW volumes across 79 Local Government Areas in Victoria, Australia (2002–2023). The hybrid models substantially improve predictive accuracy over baseline XGBoost, providing robust forecasts at the local government scale. SHapley Additive exPlanations (SHAP) identify population size, distance to the central business district and land area as the key drivers of waste generation, highlighting the influence of urban form and spatial structure on MSW pressures. The model is then used to project MSW volumes for 2026, 2031 and 2036, indicating marked increases relative to 2023 and revealing which LGAs are likely to experience the greatest growth in waste. These results underline the need for forward-looking infrastructure planning and targeted policy interventions and demonstrate how advanced ensemble models can support more evidence-based and sustainable urban waste management strategies.