MODELING OF THE DISTRIBUTION OF COVID-19 CASES IN 5 DISTRICTS IN BANDUNG CITY USING THE VECTOR AUTOREGRESSIVE MOVING-AVERAGE (VARMA) TIME SERIES MODEL WITH A FOCUS ON THE STATIONARITY AND INVERTIBILITY OF THE VARMA MODEL

• Main Author: Hasfi Narendra, Adri
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49718
• Institution: Institut Teknologi Bandung

Time series modeling is a form of mathematical modeling that guesses a value or data at a future time based on previous data. The time series model that will be used for this research is the Vector Autoregressive Moving-Average (VARMA) model, which is a multivariate form of the Autoregressive Moving-Average (ARMA) time series model. The ARMA model is a time series model of a stationary (weak) stochastic process, this model describes data based on 2 polynomials, namely the polynomial for the autoregression (AR) process and the moving average (MA) process. Because the ARMA model is a univariate model, it does not involve dependence on other variables. The VARMA model is a multivariate form of the ARMA model with each variable of the VARMA model being shaped as a vector and the coefficients as a matrix. The VARMA model has several advantages over the ARMA model, namely the VARMA model can contain more variables, the VARMA model can also capture the dependence effect in each variable used. So that in analyzing variables that have dependence on one another, the VARMA model has better accuracy and precision than the ARMA model. The disadvantage of the VARMA model is that the modeling process is more complicated than the ARMA model. The VARMA model also has more assumption requirements than the ARMA model, some of these assumptions are that the model must be stationary and invertible. In addition, in the ARMA model the variables used must be stationary, while in the VARMA model, not all variables used must be stationary. In this research, a case study of COVID-19 was carried out in 5 sub-districts in the city of Bandung. From the COVID-19 case data, a VARMA model will be created and also analyzed whether the VARMA model is stationary and invertible. If the VARMA model is not stationary and invertible, then the order of variable differentiation will be tweaked and then seen its effect on the stationary and invertible properties of the model. And if the VARMA model is stationary and invertible, a non-stationary or invertible VARMA model will be created and then its accuracy will be compared. In the research, it will also be tested whether there are variables that can be discarded.