In this study, we introduce an advanced neural network-based approach to solving optimal control problems within the Susceptible-Exposed-Infectious-Quarantined-Recovered-Susceptible (SEIQRS) epidemic model. By leveraging deep learning techniques, our method encodes control functions within neural networks, enabling efficient handling of complex epidemiological control challenges. We first establish the existence of an optimal control solution in a general framework. Then, we employ neural networks as an approximation method for control strategies, utilizing the adjoint state problem and gradient descent to iteratively refine control values through parameter adjustments in the neural network. Our approach provides novel insights into epidemic forecasting and control strategies, demonstrating the superiority of neural network-based methods over traditional direct collection approaches. The results highlight enhanced adaptability and efficiency in epidemic management, opening new perspectives for real-time decision-making and public health interventions.
Azoua et al. (Tue,) studied this question.