Applications of Artificial Intelligence in Infectious Disease Surveillance and Outbreak Predictions: Machine Learning Approaches and Public Health Implications

Background: Infectious diseases still cause millions of deaths every year, and traditional surveillance systems are often slow, miss cases, and delay outbreak detection. Machine learning (ML) offers new ways to monitor diseases and predict outbreaks faster using data like cases, weather, and movement. Objective: The aim of this review is to summarize ML approaches used for surveillance and prediction of infectious disease outbreak. Methods: This systematic review followed PRISMA 2020 guidelines. We searched PubMed/MEDLINE, Embase, Scopus, Web of Science, and Google Scholar from 2020, to the most recent records. We included only original peer-reviewed studies in English that used ML for population-level surveillance or outbreak prediction in human infectious diseases, with clear performance metrics and public health discussion. Two reviewers screened titles/abstracts and full texts independently. Risk of bias was checked with PROBAST. Narrative synthesis was used because studies varied in several aspect. Results: From thousands of records, 15 high-quality studies were included using highlighted criteria. They focused on dengue and respiratory viruses like COVID-19 and influenza. Tree-based models (XGBoost, Random Forest) were used most (80%), Deep learning (LSTM, Transformer) appeared in 47% of studies for time-series monitoring of respiratory viruses, with good accuracy. Hybrid/ensemble models often performed best. ML helped early warnings (5–14 days lead time) and worked with real data from several countries. Surveillance was improved in 12 studies, and prediction was strong in all 15 studies included. Public health benefits included better resource use and integration into national systems, but challenges were data bias, privacy, and equity. Conclusion: Machine learning showed usefulness relatively accurate and useful for faster surveillance and outbreak prediction of infectious diseases. More work is needed on real-world testing, reducing bias, and making models easy to understand so they can be used fairly everywhere.