A review of AI/ML approaches in wastewater surveillance advancement

Wastewater-based epidemiology (WBE) has emerged as a powerful tool for early detection and monitoring of infectious diseases, particularly during pandemics such as COVID-19. This study systematically evaluates the application of artificial intelligence (AI) and machine learning (ML) models in WBE over the past five years, focusing on their effectiveness in pathogen detection and disease trend forecasting. Various supervised, unsupervised, deep learning, and time-series models were compared based on their predictive accuracy, scalability, interpretability, computational demands, and real-time feasibility. Comparative analysis showed that Random Forest (RF) achieved R² values of 0.80 and Root Mean Square Error (RMSE) 0.54 for COVID-19 trend forecasting, outperforming linear regression. Support Vector Machines (SVM) improved pathogen classification accuracy by ∼20% compared with traditional analytical techniques. Artificial Neural Networks (ANN) estimated pathogen prevalence with R = 0.81-0.92 and mean squared, while Long Short-Term Memory (LSTM) networks achieved R² ≈ 0.81 (test) and 0.94 (train) for multi-community forecasting. Time-series machine learning models (TSML) frameworks consistently produced lower RMSE and Mean Absolute Error (MAE) values than ARIMAX models, confirming their real-time prediction power. Unsupervised models like K-means clustering supported outbreak pattern identification, when labeled data were limited. Additionally, a decision-support framework was proposed to guide model selection based on prediction objectives, data type, and temporal dependencies. The findings emphasize the importance of integrating hybrid modeling approaches and environmental metadata to enhance WBE systems, and they offer a foundation for real-time, adaptive surveillance strategies.