COMBINING FORECASTS OF ARIMA AND EXPONENTIAL SMOOTHING MODELS

Dengue is one of the prominent infections in the world which has significant number of morbidity and mortality rates. In Sri Lanka, the growth and the spread of the dengue has increased stating from 1960. There is a necessity of implementing controlling actions towards the dengue disease. In this regard, effective prediction models can be used. The aim of this study is to predict dengue cases in Colombo, Sri Lanka using autoregressive moving average method and exponential smoothing technique. Data consist of monthly reported dengue cases in Colombo district from January 2010 to November 2017. Data from 2010 to 2015 have used for model building and rest of the data for model validation. The selected best model for the Colombo is ARIMA(1, 1, 2)(1, 0, 0)(12). The best exponential smoothing model for Colombo consists of multiplicative error, multiplicative seasonality and additive structure for trend. To improve the forecasts generated by the two models and to remove some poor performances, several combined forecasts methods were applied. The performances of combined forecasts were tested using RMSE and MAPE measures. The OLS regression method was the best combining forecast method to predict monthly dengue cases in Colombo, Sri Lanka. The forecasted values of the combined fit will be useful in taking actions towards controlling the dengue cases in Colombo, Sri Lanka.