Neural Network Autoregressive Model for Forecasting Malaysia Under-5 Mortality

Under-five mortality is a key point of kid prosperity in which most countries have discussed on. In 2019, around 5.2 million children died each year throughout the world. Despite these countries having a high number of deaths, the lack of data makes it difficult to get accurate estimations. Moreover, this early childhood mortality is still high and has turned into a huge problem in some developing countries. Thus, this study aims to study the trend pattern and develop forecasting models to forecast future trends of under-five mortality in Malaysia by gender. The yearly under-five mortality rates (U5MR) of 41 years (1980�2020) in Malaysia were analysed using Neural Network Autoregressive (NNAR). The result of the NNAR was then compared with the Box-Jenkins Methodology (ARIMA model) result. It was found that the U5MR in Malaysia fluctuated from year to year with a slowly decreasing trend pattern for both genders, with males having a higher rate than females. Moreover, the result from the NNAR model provides a more accurate forecast compared to the ARIMA model for both genders with the lowest root mean square error (RMSE) and mean absolute percentage error (MAPE) value. The future trend increased slightly and the forecast trend for the male was higher than the female population. The result of this study could become a reference for other developed and developing countries. It could also become an indicator for human resource management and health care allocation planning. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.