Machine Learning-Based Prediction of Infectious Healthcare Waste Generation: A Multi-Clinic Study of 24 Clinics at the Military Medical Academy

Effective management of infectious healthcare waste at the Military Medical Academy (VMA) depends on reliable forecasting in order to ensure adequate treatment capacity (e.g., sterilization facilities), optimize logistics, maintain regulatory compliance, and minimize environmental impact. However, conventional statistical approaches often struggle to capture the complex and heterogeneous patterns of waste generation ob-served across clinical departments with different medical specializations. The aim of this study is to develop and comparatively evaluate six models for predict-ing annual infectious waste generation across 24 clinical departments of the Military Medical Academy in Belgrade, Serbia. The analysis is based on an 11-year real-world panel dataset (2011–2021), which is further used to produce forecasts for the period 2022–2031. The modeling framework includes both traditional statistical methods (OLS, Ridge, and Lasso regression) and machine learning techniques (Random Forest, Gradient Boosting, and multilayer perceptron). Model performance is assessed using k-fold cross-validation and standard evaluation metrics (RMSE, MAE, and R²). The results indicate that machine learning models, particularly Gradient Boosting and Random Forest, achieve better predictive performance compared to traditional approaches. In addition, the analysis of feature importance provides insight into key factors influencing waste generation, which may support more informed planning and resource allocation within hospital systems. Although the findings are based on data from a single hospital complex, they offer a useful empirical basis for understanding and forecasting infectious healthcare waste in large, multi-department healthcare institutions.