Transformer-based neural networks for epidemiology

Project Level: Honours, PhD

In this project, we leverage the power of state-of-the-art transformer networks in predicting time-series data to forecast the number of patients for epidemic diseases like influenza and COVID-19. The project includes using interpretation methods to assess the role of each predictor in the prediction made. We also analyse instances of high error in prediction to identify the conditions in which a machine learning method might do a poor job in predicting the number of patients.