Optimizing biomedical waste generation modeling using quantum machine learning and economic development indicators

Reliable biomedical waste (BMW) prediction is essential for designing efficient waste management systems that protect sustainable cities from health risks. Machine learning (ML) modeling provides efficient and accurate systems, which aid in maximizing management operations. This study analyzes three standalone ML models, namely support vector regression (SVR), narrow neural networks (N-NN), and optimized SVR with quantum behavior particle swarm (QPSO-SVR) for predicting BMW generation rates. The study aimed to assess the data reliability and applicability of ML models for supporting data-driven strategies in waste management planning and public health policy. Feature engineering was used to determine the input variables, and model performance was evaluated using statistical indices aided by 2D visualizations. The prediction outcome indicated that N-NN achieved the highest predictive accuracy (95 %), outperforming SVR and QPSO-SVR (both 91 %). The testing phase further revealed an increased performance with SVR recording the lowest mean squared error (MSE = 0.0108(kg/day)), followed by QPSO-SVR (MSE = 0.0111(kg/day)), indicating strong generalization. Data reliability was verified using Augmented Dickey-Fuller (ADF), Phillips-Perron (PP), and Jarque-Bera (JB) tests. ADF and PP tests confirmed stationarity, and JB confirmed partial normality. The results demonstrate the ability of ML models to enhance forecasting accuracy. This will support informed decision-making in sustainable waste management and public health protection. Future work could focus on developing hybrid models with advanced data integration to utilize complementary strengths and improve the accuracy and robustness of real-time predictions.