AI-Enhanced SEIR Modelling for Adaptive Control of Diphtheria Outbreaks in Delta State, Nigeria

Mathematical models like the Susceptible-Exposed-Infectious-Recovered (SEIR) framework are vital for epidemic control but are often limited by static parameters in dynamic settings. This study develops a hybrid framework that integrates Pontryagin's Maximum Principle for optimal control with artificial intelligence (AI) components—including Long Short-Term Memory (LSTM) networks for forecasting and a Genetic Algorithm (GA) for real-time parameter adaptation—to transform a traditional SEIR model into an adaptive decision-support system. Applied to diphtheria outbreak management as a case study in Delta State, Nigeria, the AI-augmented model was evaluated using synthetic data derived from Nigeria Centre for Disease Control parameters. It reduced forecast error by 80% (MAPE: 4.6% vs. 22.8%), peak infections by 66.5%, and improved intervention cost-efficiency compared to the optimized control-only model. The findings suggest that AI-mechanistic modelling provides a scalable methodological framework for precision public health, especially in resource-limited settings, through improvements in predictive performance, resiliency, and cost-effectiveness.