Artificial intelligence and machine learning in predicting food hunger and severity: A comprehensive review

Food insecurity has been identified as one of the biggest twenty-first century humanitarian and public health emergencies. As of 2023, there are an estimated 733 million people facing food insecurity globally. The nature of food insecurity, which is shaped by the interwoven drivers of climate variability, armed conflict, economic shocks, structural poverty, etc., has historically surprised traditional prediction models. Artificial intelligence (AI) and machine learning (ML) have proven to be methodological tools with great potential to combine high-dimensional, heterogeneous data streams and produce early alerts of food insecurity onset, spatial mapping of food insecurity severity, and actionable intelligence for humanitarian response. This review aims to summarize the AI and ML use for worldwide food insecurity forecasting, including supervised methods like random forests, gradient boosting, and deep neural networks; clustering of unsupervised food insecurity vulnerabilities; the extraction of crisis signals from social media and news corpora using natural language processing; and the assessment of crop health through computer vision techniques applied to satellite and drone images. The operational platforms are critically examined about algorithmic foundations and predictive quality, such as the World Food Programme's Hunger Map Live, FEWS NET, FAO's Global Information and Early Warning System (GIEWS), and the Integrated Food Security Phase Classification (IPC). Persistent issues such as limited availability of data in conflict-affected areas, the lack of interpretability in models, algorithmic bias towards marginalized communities, and the digital infrastructure gap in low-income countries are further explored. The review also outlines a research agenda focused on federated learning architectures, explainable AI frameworks, and integrating climate non-stationarity into predictive models, aiming to bring algorithmic progress to policy action that will save lives and distribute benefits equitably.