Prophet forecasting and temporal modeling of Covid-19 cases in the Philippines

This paper aims to make use of machine learning models to forecast the trend of COVID-19 in the Philippines to approximate when surges might occur and for the country to be better prepared for the next wave. This study uses 5 different forecasting techniques: Naïve method, Simple Exponential Smoothing (SES), Double Exponential Smoothing (Holt’s Linear Trend Method), Auto Regressive Integrated Moving Average (ARIMA) and Prophet Forecasting Model. Compared to the four other techniques, the Prophet Method is a fairly new forecasting method developed in 2017 by Sean J. Taylor and Ben Letham. The model performances were compared with the use of Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and with the addition of Akaike’s Information Criterion (AIC) for ARIMA. For the study, the dataset used was obtained from Our World In Data (OWID) website and contained the number of daily confirmed cases and deaths in the Philippines. The training set was used to forecast future values for the 5 methods, each forecasted result was then compared to their test data to measure model performance against each other. Results showed that Prophet outperformed all other methods with it having the lowest RMSE and MAPE