Nonseasonal ARIMA Modeling and Forecasting of Malaria Cases in Children under Five in Edum Banso Sub-district of Ghana

Abstract

Approximately sixty-nine percent of deaths globally resulting from malaria infection occur in children under five years of age [1]. National control and prevention strategies will be greatly enhanced through better ability to forecast future trends in the disease incidence. This study sought to develop a five year forecasting model for malaria incidence among children under the age of five in the Edum Banso sub- district of Ghana. A secondary monthly malaria incidence record data spanning the period of 2008-2013 was used to develop the model formulated. The Box and Jenkins ARIMA methodology was employed to obtain the best fitting model by comparing the normalized BIC, MAE and Stationary-R Square values, and Q -plots of residuals of the suggested models. Comparative analysis showed that ARIMA (1, 1, 2) has best performance. The five year forecast showed a linear trend angled diagonally up. This means malaria cases in children under five in the Edum Banso sub-district of Ghana will increase steadily over the next five years (2013-2018). The high value of the Stationary-R Square 0.792 or 79.2% is an indication that 79.2% of the variation in the malaria cases can be explained by the data. The study suggests that the department of disease control and prevention in Ghana uses ARIMA (1, 1, 2) model to optimize malaria prevention by providing estimates on malaria morbidity among children under the age of five  in the Edum Banso sub-district.