AI Solutions for Predicting Epidemic Outbreaks

Sammanfattning

This thesis explores how AI (Artificial Intelligence) models can used to predict the spread of infectious diseases, with a focus on COVID-19. The study evaluated different forecasting models like LSTM, GRU, XGBoost, ARIMA, and Prophet on Swedish healthcare data. These models trained using three strategies like fixed, sliding, and expanding windows to understand forecasting results. The analysis showed that XGBoost and LSTM, were able to predict case trends accurately. The models evaluated using parametric metrics like MAE and RMSE. Among all approaches, XGBoost with a sliding window gave the best results. This study concludes that the importance of better choice of the model and training method for forecasting outbreaks. These findings can help healthcare systems to take decision very quickly.