Background: The increasing burden of antimicrobial resistance (AMR) in hospitals necessitates proactive strategies for surveillance and intervention. Inspired by predictive models used in traffic forecasting, we propose a novel machine learning (ML)-based algorithm that integrates multiple hospital parameters—antibiotic consumption patterns, resistance rates, patient admissions, and bed occupancy trends—to predict AMR evolution and guide antimicrobial stewardship programs.

Methods: The algorithm utilizes long short-term memory (LSTM) networks and gradient boosting models (e.g., XGBoost) to analyze temporal patterns in hospital antibiotic usage and AMR incidence. A dataset including historical antibiograms, hospital admission rates, antibiotic prescribing data, and bed occupancy rates was used for training. The model employs autoregressive integrated moving average (ARIMA) and reinforcement learning techniques to enhance predictive accuracy. Performance was assessed using mean absolute percentage error (MAPE) and root mean squared error (RMSE) against real-world AMR data from hospital surveillance reports.

Results: The LSTM-based model demonstrated superior accuracy in predicting resistance trends, identifying early warning signals for carbapenem-resistant Enterobacterales (CRE) and methicillin-resistant Staphylococcus aureus (MRSA) outbreaks. The algorithm outperformed traditional statistical models, enabling real-time optimization of antibiotic prescriptions and resource allocation.

Conclusion: AI-driven predictive analytics, originally developed for traffic modeling, can be repurposed for hospital AMR surveillance. Implementing such models in clinical workflows could improve infection control strategies and reduce resistance development. Future work should focus on real-time deployment and multi-hospital validation.