Predicting malaria incidence in Kenya using the ARIMA and SARIMA models

Abstract

Malaria is considered a public health challenge across the world. Approximately 40 percent of the population of the world is at risk of malaria. In this study we will employ time series analysis models to predict malatia incidence and it will also use climate variables such as temperature and rainfall as exogenous inputs. Future malaria incidences will be projected based on detennined trend patterns. The Auto-regressive integrated moving average (ARIMA) and Seasonal Auto-regressive integrated (SARIMA) models were used in this study to predict and forecast monthly malaria incidence (the spread of malaria in Kenya). Considering the results, ARIMA (0, 1, 0) appeared fit for forecasting monthly malaria incidence in Kenya further on the SARIMA model was used to compare which model had the best results and to remove seasonality from the data the best fit for SARIMA model was (0, 1 ,0) (0, 1 ,0)[ 12]. The models that usually give slightly better results are the ones that have the lowest AICc values. In addition to that a regression analysis was carried out to detennine the effects of rainfall and temperature on malaria incidence in Kenya. The variables have different orders; we estimate a V AR regression because V AR regression enables us to dynamically measure variables with combination I different order. The results obtained from the regression analysis indicate that temperature has no significant impact on the number of malaria cases however rainfall has a significant impact.