A hybrid model for forecasting communicable diseases in Maldives

• Main Authors: Rad, Babak Bashari, Shareef, Ali Aseel, Thiruchelvam, Vinesh, Afshar, Andia, Bamiah, Mervat Adib
• Format: Article
• Published: Taylor's University 2018
• Subjects: R Medicine
• Online Access: http://eprints.um.edu.my/20999/; http://jestec.taylors.edu.my/Special%20Issue%20ICCSIT%202018/ICCSIT18_01.pdf

The Maldives is an island nation and the islands are scattered over 26 atolls. The government of Maldives is trying to improve health services in the country and improve the accessibility of services throughout the country at the peripheral levels. The healthcare industry collects a large amount of healthcare information, which contains several patterns, such as outbreaks of diseases. However, this data frequently goes unexploited. Accurate forecasting using this past data could help healthcare managers in taking appropriate decisions especially in implementing preventing measures. Due to the geographical nature of Maldives, it is difficult to implement preventive measures in case of an outbreak. There is no single approach to be used for health forecasting; thus, various methods have been used to specific health conditions or healthcare resources. Healthcare comprises of both complex linear and nonlinear patterns, which can affect the forecasting accuracy if only linear models or neural networks are used. In this research, a hybrid of the ARIMA model and Neural Network has been proposed to forecast healthcare data. A dataset comprising of 10 diseases including unique cases reported for each disease, between the years 2012 and 2016 have been used in this research. It was found that the proposed model performed well on 7 out of the 10 diseases.