Forecasting Ethiopia's Neonatal Mortality Rate Toward the Sustainable Development Goal Target 3.2 Using Statistical and Deep Learning Models: A Time-Series Forecasting Study

Background and aims: A high neonatal mortality rate (NMR) reflects underlying health inequalities, the quality of maternal and newborn care, and overall healthcare system performance. Despite substantial progress, Ethiopia remains off track to achieve Sustainable Development Goal (SDG) 3.2, which aims to reduce NMR to at least 12/1000 by 2030. This study compared statistical and deep learning models to forecast Ethiopia's NMR through 2030, assess the progress towards the SDG target, and provide evidence to inform policy.

Methods

Annual NMR data were obtained from the WHO Health Inequality Data Repository. Autoregressive integrated moving average (ARIMA), long short-term memory (LSTM), multi-step LSTM, and a hybrid were developed using Python. For the deep learning models, the data were normalised using MinMax scaling and transformed into sequential time-series windows. Stationarity was assessed using rolling statistics and the Augmented Dickey-Fuller test, while ARIMA parameters were selected using autocorrelation and partial autocorrelation and diagnostic criteria. Model performance was evaluated on the test dataset using mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), symmetric MAPE (sMAPE), and mean absolute scaled error (MASE). The best-performing model was used to forecast NMR from 2024 to 2030.