Accuracy of short - term forecasting for occurrence of dengue diseases using time series analysis, 2008 - 2013

Abstract

Disease forecasting is useful for effective planning of prevention and control program. Time series analysis is one of quantitative methods applied in disease forecasting. Objective of this study is to assess accuracy of forecasting of dengue diseases using the method. Dengue data reported in national notifiable disease reporting system during 2003 - 2013 were analyzed to develop forecasting models. Errors of one-, two-, three-, and six-month forecasting from two methods, i.e. exponential smoothing and Box - Jenkins methods, during 2008 - 2013 were compared. Mean absolute percent error of the forecasts from exponential smoothing at one-, two-, three-, and six-month were 16%, 33%, 47% and 54% respectively, while those from Box – Jenkins methods were 16%, 31%, 40% and 60% respectively. The findings suggest that, to achieve accurate forecasting for dengue diseases with error fewer than 20%, one - month forecasting is recommended.