FORECASTING THE SPREAD OF COVID-19 WITH PROPHET MODEL USING BELGIUM DATASET

Abstract

The Coronavirus pandemic emerged as the deadliest in recent times, originating from China in 2019 and affecting the entire globe. The data used in this work is obtained from a publicly available official Belgian website. Model parameters of ARIMA were selected, and the optimized parameters were used to forecast the COVID-19 cases. Daily time series data from April 2020 to July 2021 of total infected cases from Belgium were collected. ARIMA forecast for 60 and 100 days was computed using the seasonal ARIMA model and PROPHET model, which resulted in the forecast with actual (observed) data and predicted data. ARIMA and PROPHET model suggested the continuing trend of infection in Belgium. The predicted ARIMA forecast model (3,0,0) (0,1,0) had MAPE 39.8, with Mean Error 0.005 and RMSE 0.54. The obtained model may be used to assess the successive waves of infection in other countries. The Prophet model outperforms ARIMA in accuracy in forecasts. The RMSE of the model was 259.7536 and MAE was 208.8522.