Prediction of Bontang City COVID-19 Data Time Series Using the Facebook Prophet Method

ABSTRACT

The increasing trend of COVID-19 cases in Bontang makes it the first order of the highest incident rate in East Kalimantan, with a value of 1161.78 cases per 100 thousand inhabitants. The purpose of this study was to predict the increase in COVID-19 cases in Bontang City with a data set of positive confirmed cases, recovered and died of COVID-19 in Bontang city. The data set used starts from March 24, 2020 - March 1, 2021, using the Facebook Prophet method, the Jupyter Notebook application, and the Python programming language. The research process consists of the data collection stage, prediction implementation stage (data preprocessing, processing, performance evaluation, dashboard creation), and analysis of the result. The prediction was performed for up to 92 days until May 5, 2021. The result shows a trend of increasing cases of covid reaching the highest positive value, the highest recovery, and highest death, respectively, of 8695, 6099, and 156 people. According to the model, the average positive predictive error (MAE) and the average positive predictive accuracy value (MAPE) are 0.17 and 17.4%, indicating the positive prediction of contracting covid has good accuracy criteria. The next evaluation for the death prediction is accounted as reasonable accuracy criteria in which MAE and MAPE are 0.27 and 27%, respectively. Lastly, the recovery prediction has MAE of 0.17 and MAPE of 17.4%, implying good accuracy criteria. The study also provides recommendations to the COVID-19 Task Force to prepare the minimum number of PCR Tests by 870 tests and increase the hospitalization occupancy by 294 to control the spreading of the Coronavirus. © 2021 IEEE.