| Summary: | This study applies Natural Language Processing (NLP) and Machine Learning (ML) to investigate global historical trends in food security. Using USAID’s Famine Early Warning Systems Network’s (FEWS NET) comprehensive reports spanning over two dozen countries, it explores prevalent dimensions such as shocks, outcomes, and coping capacities, offering insights into long-term food security conditions. Results highlight the prevalence of market and climate impacts across the countries and period considered. Based on results from the topic classification, ML models were applied to determine the most important factors that predict food insecurity. The analysis confirmed market shocks as the main predictors of food insecurity globally. The approach demonstrates the potential for extracting valuable insights from narrative sources that can support decision-making and strategic planning. This integrated approach not only enhances understanding of food security but also presents a versatile tool applicable beyond the context of humanitarian aid.
|
|---|