Forecasting Measles Cases in Ethiopia using Neuro-Fuzzy Systems

Measles is still one of the deadly diseases that terrify the world. Making accurate decisions is necessary in organizing an effective fight against the disease. Forecasting is a very important tool for planning future needs of the health services including the vaccine stocks. Forecasting methods will help to watch over the critical stock levels of the medication as well as their expiration dates and perhaps take the necessary action to ensure a health stock flow. Governments and health institutions estimate the measles vaccine requirements using certain equations generally based on the size of the target population and the past consumption records. There are several studies that have examined the measles forecasting and conducted a vaccine requirement assessment. In this study genetic algorithm (GA) based trained recurrent fuzzy neural network (RFNN) and Adaptive neuro-fuzzy inference system (ANFIS) used to forecast the monthly measles cases in Ethiopia. The Ethiopia measles data was extracted from the World Health Organization Measles and Rubella Surveillance Data, which covers the period from January 2011 to December 2017. Out of total monthly measles cases, 80% were used for training, and 20% were chosen for testing. GA based trained RFNN shows better performance compared to ANFIS results.