Abstract:
A crucial focus of public health surveillance systems is to provide reliable forecasts of epidemiological time series. This work utilized data collected through a national public health surveillance system in Thailand to evaluate and compare the performance of a seasonal autoregressive integrated moving average and an Integer generalized autoregressive conditionally heteroscedastic model for modeling and forecasting case occurrence of dengue. The comparison uses weekly reported cases of dengue hemorrhagic fever in Amnat Charoen province Thailand, from January 1st, 2006, to October 7th, 2017 (612 weeks). The results from the in-sample evaluation using the root mean square error and mean absolute error as well as a visual inspection of predicted values show that the two approaches are adequate tools for use in epidemiological surveillance as there is no significant difference in their forecast accuracy for in-sample performance. The incorporation of the weather variables improves the predictive performance of the models and from the model coefficients the study findings reveal that there is a positive relationship between temperature and rainfall and the occurrence of dengue. Overall, the findings in this study support the usefulness of the two approaches as effective tools practitioners can utilize for monitoring and for providing early warning signals of potential outbreaks of epidemics.