AI-Based Approaches for Food Security: An In-Depth Study on Predictive Models and Resilience

This work integrates Bayesian Networks (BN) with Shannon's Information Theory to provide a fresh method of evaluating and projecting world food security. This hybrid method, unlike conventional statistical models, allows probabilistic reasoning and efficient uncertainty quantification. Using the 2019 Global Food Security Index (GFSI) dataset, the paper painstakingly assesses the influence of several elements on food security, especially under extreme conditions. Continuous data was turned into meaningful variables for BN modeling by means of BayesiaLab for data discretization and R2-GenOpt for category optimization. The built BN model found that the GFSI target node was significantly influenced by about 24% of variables. With an absolute Gini value of 70.77% and a relative Gini index of 99.96%, the model shown strong predictive performance, thereby stressing its accuracy in predicting food security results. Predictive simulations confirmed even more the capacity of the model to evaluate resilience policies. By means of artificial intelligence-driven decision-making, the study offers a disciplined framework for companies and legislators to improve food security risk management. These results highlight BN models' promise for strategic planning and adaptive policy development. Future studies could improve these approaches by adding real-time data and extending predictive powers to handle issues related to world food security holistically.