Neonatal Mortality Forecasting in Somalia: A Comparative Time-Series Analysis of Single and Hybrid Models

This repository contains a reproducible time-series dataset and analytical outputs for neonatal mortality forecasting in Somalia. The dataset includes annual neonatal mortality rate (NMR) data for Somalia covering the period 1983–2023, obtained from the World Bank Open Data repository (Indicator: SH.DYN.NMRT). The data were cleaned, validated, and structured for time-series analysis. The study applies a comprehensive multi-model forecasting framework, including ARIMA, ETS, TBATS, Theta, ARFIMA, and Neural Network Autoregression (NNAR), along with hybrid model combinations. Model performance was evaluated using standard accuracy metrics (MAPE, sMAPE, and Theil’s U). The archive includes:- Processed mortality dataset (CSV format)- Forecast results for 2024–2030 with prediction intervals (80% and 95%)- Model accuracy metrics for all models- R scripts for data preprocessing, stationarity testing (PP and KPSS), model fitting, and forecasting- Supplementary materials including tables, figures, and methodological documentation- README file explaining dataset structure, variables, and usage Forecast results indicate that neonatal mortality in Somalia is expected to stabilize at approximately 35 deaths per 1,000 live births over the forecast period, with increasing uncertainty over time. This archive ensures transparency, reproducibility, and open scientific access to the full analytical workflow. License: Creative Commons Attribution 4.0 (CC BY 4.0)