Page 93 - programme book

ST-027
Spatio-temporal Analysis of COVID-19 Spread in Selangor

Saidatul Nurfarahin Muhamad Yusof 1, a) , Nur Haizum Abd Rahman 1, b) , Iszuanie Syafidza
Che Ilias 2, c) , Kathiresan Gopal 2, d) and Noraishah Mohammad Sham 3, e)

1 Department of Mathematics and Statistics, Faculty of Science, Universiti Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
2 Institute for Mathematical Research, Universiti Putra Malaysia,
43400 UPM Serdang, Selangor, Malaysia.
3 Environmental Health Research Centre, Institute for Medical Research, 40170 Shah Alam,
Selangor, Malaysia.
.
a) 197088@student.upm.edu.my
b) Corresponding author: nurhaizum_ar@upm.edu.my
c) iszuanie@upm.edu.my
d) kathiresan@upm.edu.my
e) noraishah.ms@moh.gov.my

Abstract. The coronavirus disease 2019 is a novel pandemic that has spread over the world, affecting
every country including Malaysia. Therefore, forecasting the number of positive COVID-19 cases is
crucial especially in Selangor where the cases are considered high. The COVID-19 data can be
modelled and forecast by using time series methods on both univariate and multivariate modelling.
The spatio-temporal model is one of the multivariate models because it is not only linked to the events
from the past time, but also has relevance to the other location. This study aims to model and compare
the forecast of COVID-19 confirmed cases by using the integrated generalized space-time
autoregressive (GSTARI) model with different weights which are uniform and inverse distance weight
and autoregressive integrated moving average (ARIMA) model in three districts which are Petaling,
Hulu Langat, and Klang. The parameter estimate was done by using the ordinary least squares (OLS)
method and the performances were compared using root mean square error (RMSE). The study showed
the GSTARI (1,1) model with uniform and inverse distance weight outperformed ARIMA model in
Petaling and Klang districts but not in Hulu Langat. However, as overall the GSTARI model still gave
the best result in forecasting since GSTARI model only required less computational run time compared
to the ARIMA model.

Keywords: Spatio-temporal, GSTAR, ARIMA, COVID-19, forecasting