Abstract Background High infant mortality rate (IMR) is a critical public health indicator. Understanding its trajectory is vital for policymaking, particularly in Somalia, where rates remain elevated. This study aimed to model and forecast Somalia’s IMR to identify the most accurate method, providing data-driven insights for public health strategy. Objectives The study aimed to formulate time series models, propose hybrid models, and rigorously compare their performance in forecasting IMR. Methods This quantitative time-series analysis utilized annual World Bank data for Somalia from 1950 to 2022. Three models were developed: Autoregressive Integrated Moving Average (ARIMA), Artificial Neural Network (ANN), and a hybrid ARIMA-ANN model. Performance was evaluated using Root Mean Square Error (RMSE) and Symmetric Mean Absolute Percentage Error (sMAPE) on a test set spanning 2013 to 2022. Results Upon achieving stationarity via first-order differencing, the ARIMA (1, 1, 3) model proved most effective. It demonstrated superior accuracy with an RMSE of 0.85, significantly outperforming the ANN (RMSE 1.04) and hybrid models (RMSE 3.01). Validated ARIMA forecasts project a continued, gradual decline in IMR from 59.02 deaths per 1,000 live births in 2023 to 40.37 in 2032. Conclusion The ARIMA model offers the most reliable framework for forecasting Somalia’s infant mortality rate based on historical data.
原文链接
分享
引用
