Forecasting Model for Pneumonia Cases in 13 Health Districts in Thailand

Abstract

The objective of this research is to evaluate a proposed forecasting model for predicting pneumonia cases across all 13 health districts in Thailand. Monthly time series data from January 2015 to December 2022, obtained from the Health Data Center of the Ministry of Public Health, were utilized. Each health district utilized a varying number of data points depending on data completeness. The data were categorized into two sets: a training dataset for model construction and a test dataset for assessing model accuracy. The forecasting models were evaluated using the symmetric mean absolute percentage error (sMAPE).

The results revealed that the combined forecasting approach using regression analysis demonstrated advantages over other methods for health districts 1 to 6, 8, 9, 10, 12, and 13. For health district 7, the most appropriate model was the decomposition model with trend and seasonal components in a multiplicative form. Similarly, for health district 11, the Box-Jenkins model was found to be the most suitable.