Forecasting HIV Incidence in Somalia Using Single and Hybrid Time Series Models: A Comparative Analysis

Abstract

Background HIV incidence is a key indicator for monitoring new infections and assessing progress in HIV prevention. In Somalia, evidence on long-term HIV incidence trends remains limited because of constraints in routine surveillance and the scarcity of forecasting studies. This study examined historical trends and forecasted HIV incidence in Somalia using individual and hybrid time series models. Methods Annual HIV incidence data for Somalia from 1990 to 2022 were obtained from the World Bank World Development Indicators database. Stationarity was assessed using the Augmented Dickey–Fuller, Phillips–Perron, and Kwiatkowski–Phillips–Schmidt–Shin tests. After second-order differencing, six individual forecasting models—ARIMA, ETS, TBATS, Theta, NNAR, and ARFIMA—and selected hybrid models were fitted. Model performance was evaluated using mean absolute percentage error, symmetric mean absolute percentage error, and Theil’s U statistic. Data were divided into training and testing periods covering 1990–2014 and 2015–2022, respectively. Results The original HIV incidence series was non-stationary and became stationary after second-order differencing. Among the individual models, ETS (M,N,N) showed the strongest forecasting performance, with the lowest error values (MAPE = 0.038%, sMAPE = 0.00038, Theil’s U = 0.005). Among the hybrid models, ETS–NNAR performed best (MAPE = 0.321%, sMAPE = 0.00321, Theil’s U = 0.040). Forecasts for 2023–2029 indicated that HIV incidence is likely to remain close to 0.10 per 1,000 uninfected population, with wider prediction uncertainty over time. Conclusions HIV incidence in Somalia declined over the long term and has recently stabilized at a low level. ETS was the best-performing individual model, while ETS–NNAR was the best-performing hybrid model. These forecasts suggest continued stability if current patterns persist, while highlighting the need for sustained HIV surveillance and prevention efforts.