A Comparative Study to Determine Optimal Models for Forecasting the Number of Patients Having Epidemiological-Surveillance Diseases in Bangkok

Abstract

The objective of this study was to determine optimal models for forecasting the population of Epidemiological-Surveillance Diseases in Bangkok. In this study we applied 5 forecasting techniques to analyze the data. These include Simple Moving Average method, Simple Exponential Smoothing method, Box-Jenkins method, Ratio-To-Trend method and Exponential Smoothing Holt-Winter method. The suitable forecasting models were chosen by considering the smallest value of MAPE. Our raw data were secondary data, which were taken from the Bureau of Epidemiology, Department of Disease Control, Ministry of Public Health. All 10 diseases data were separated on a monthly basis between from January 2008 and December 2012 and between January 2005 and December 2012. Through the comparative study of the 5 forecasting methods, the results showed that the Simple Moving Average method is the most appropriate forecasting method for almost all time-series data with no trend and seasonal characteristics. On the other hand, the Exponential Smoothing Holt-Winters is the most appropriate forecasting method for almost all time-series data exhibiting trend and seasonal characteristics.