Data-driven approach to early estimate the inflection points of infected cases in Saxony, Germany

We develop a data-driven forecasting model of COVID-19 for Saxony using autoregressive integrated moving average (ARIMA) and Holt’s models in the presence and absence of seasonal parameters. Owing to a daily-updated data curation facility, we employ a version control of daily-updated data for Saxony which serve as training data of seasonal parameter to forecast up to 4 horizons. We find that this method is capable of immediately estimating inflection points after a turning point is present. The results are also compatible with the counties of Saxony. We also tried to use multiple datasets (including infection, death, recovery, vaccination data) to train a random forest machine learning model. The preliminary result looks promising and further exploration will be done.