ESR’s data science team has developed an AI-based hospitalisation prediction system to report weekly forecasting of rates of hospitalization. The system is designed to aid the Ministry of Health in the effective allocation of healthcare resources.
The prediction system uses a state-of-the-art Transformer deep-learning architecture as the basis for the forecast model. This allows the system to capture intricate data patterns and temporal relationships, resulting in significantly improved prediction accuracy over traditional time series methods.
This model is focused on predicting the hospitalisation burden accurately in the short-term, rather than other tools that can characterise a longer-term scenario or evaluate clinical and epidemiological predictors. The system also opens further opportunities for strengthening pandemic response strategies for future outbreaks.