An analytical framework for medical waste forecasting using machine learning: Paving the path toward zero waste in healthcare

This study presents a robust framework for accurately forecasting medical waste (MW) generation in healthcare settings, addressing escalating environmental and public health concerns. Using machine learning (ML) models, including Linear Regression (LR), Support Vector Regression (SVR), Long Short-Term Memory (LSTM), and Random Forest (RF), the research integrates critical patient-related variables, such as indoor and outdoor patient counts, surgeries, and deaths. Based on eight years of comprehensive data (2016–2023) from the medical colleges in Rajshahi, Bangladesh, the LR model showed the most reliable outcomes, achieving R2 values between 0.91 and 0.95 for all waste categories and outperforming others with mean absolute errors (MAE) as low as 1.39 kg for recyclable waste. Projections reveal a significant 50 % surge in general waste by 2030, reaching approximately 987.75 tons, while infectious waste will grow from 66.54 to 735.23 tons. Monte Carlo simulations quantified variability across waste categories, demonstrating fluctuations within ±5% of mean values, ensuring robustness. Finally, the research represents a novel policy recommendation for zero waste generation at the study location, which advocates for category-specific MW management strategies, aligning with Sustainable Development Goals (SDGs) 3 (Good Health and Well-being), 12 (Responsible Consumption and Production), and 13 (Climate Action). This work introduces a novel precedent for incorporating artificial intelligence into healthcare operations to attain zero waste generation while tackling public health and environmental issues. Thus, this work offers an innovative methodology that integrates machine learning to link patient data with waste generation, providing practical insights for resource optimization and operational planning.