SPATIO-TEMPORAL MODEL FOR PREDICTING COVID19 CASES IN INDONESIA

Abstract

Objective: Spatio-temporal modelling is a method used for data which has spatial (area) and temporal (time) property. Confirmed cases of Covid19 in each province Indonesia were recorded from March 2nd to September 15th, 2020. The spatio-temporal model in this study are split into two parts which are ARIMA(p,d,q) for the temporal pattern and Bayesian Poisson regression to explain the spatial pattern.Method: Data for the study was obtained from Data Repository of Indonesian National Board for Disaster Management - Indonesia Task Force for Covid-19 Rapid Response (Gugus tugas Percepatan penangana Covid19) official website which are an opened source data. The Rstudio, Arcgis and excel was used to carry out the statistical analysis involved in the investigation. In the temporal analysis, data was assumed to have an increasing trend and to create a stationary series, an integrated method was conducted. Box-Jenskin and Ljung-Box method was taken in parameter estimation and model identification process. For the spatial analysis, a Bayesian Poisson Regression is fitted to the dataset with Metropolis algorithm.Result: Model IMA(1,1), in general, can explain he increasing trend in the Covid19 confirmed cases in Indonesia. This model can define that the case number at the particular time is affected by the moving average at lag-1. Meaanwhile, a Bayesian Poisson Regression can elaborate spatial pattern in the data. The fitted model shows that the confirmed cases at particular province is also affected by the population density at those provinces. As there are some limitation in the data and method applied in the study, further analysis and research are needed.