Epidemiological Spatiotemporal Data Exploration and Prediction

This paper addresses epidemiological spatiotemporal datasets such as those reporting the number of cases of infectious diseases over time and by geographical location. It studies methods for exploratory data analysis and for prediction of future cases based on prior data. It emphasizes methods that provide explainable predictions, such as those based on rules and decision trees. These methods are studied in the context of a recently published dataset of weekly Chickenpox cases in Hungarian counties over a 10-year period. As noted in prior work, this dataset exhibits several features, such as seasonality and heteroskedasticity, that make the prediction task especially challenging. This paper describes some results of an experimental study of both the exploratory and predictive aspects.