Clinical and Medical Engineering Live
ISSN: 2997-612X | DOI: 10.64030/2997-612X
Ismail Mahamoud yousif, Suhaib Mohamed kahie and Abdisalam Hassan Muse
Malaria remains a major public health challenge in Somalia, where evidence-based decision-making is complicated by the need to balance predictive accuracy with epidemiological interpretability. This study evaluates a dual-approach modeling framework by comparing the forecasting performance of eight statistical time-series models with a mechanistic Susceptible–InfectedRecovered–Susceptible (SIRS) model using annual malaria incidence data from 2000 to 2022. The objective was to assess both outof-sample predictive accuracy and the ability of each approach to generate actionable epidemiological insight. A clear trade-off between accuracy and interpretability was observed. Among the statistical models, the Autoregressive Fractionally Integrated Moving Average (ARFIMA) model demonstrated superior forecasting performance, achieving a Mean Absolute Percentage Error (MAPE) of 7.44% on a hold-out test set. In contrast, the SIRS model exhibited lower predictive accuracy (MAPE = 24.15%) but provided important mechanistic insight into malaria transmission dynamics. The model estimated a basic reproduction number (Râ??) of 1.19 and identified transmission and recovery rates as key parameters influencing epidemic behavior. These findings indicate that statistical and mechanistic models serve complementary roles rather than competing purposes. The results support the use of accurate statistical models, such as ARFIMA, for short-term operational forecasting and resource allocation, alongside mechanistic models, such as SIRS, for long-term strategic planning and intervention prioritization. This integrated framework offers a more robust and context-appropriate evidence base for malaria control and planning in Somalia.