An AI-Based Early Warning Algorithm for Food Security in Somalia: A Hybrid Time-Series Forecasting Framework

Somalia continues to suffer from chronic and severe food insecurity, intensified by the combined impacts of climate change, political instability, and prolonged armed conflict. The highly nonlinear and interdependent nature of socio-economic and environmental variables poses significant challenges for traditional linear forecasting approaches, which often fail to capture the complex dynamics underlying food shortages. To overcome these limitations, this study develops an artificial-intelligence-based hybrid prediction framework that integrates Long Short-Term Memory (LSTM) networks with the eXtreme Gradient Boosting (XGBoost) algorithm. The proposed model predicts food shortage risks 3-6 months in advance by leveraging the temporal dependency modeling capability of LSTM and the nonlinear feature learning strength of XGBoost. By jointly exploiting temporal patterns and multi-source data relationships, the framework enables more accurate and robust forecasting. Experimental results demonstrate that the LSTM-XGBoost fusion model achieves superior predictive accuracy and earlier warning capability compared with conventional statistical methods, thereby providing a reliable decision-support tool for humanitarian response and early intervention planning.