Highlights • Compartmental models with time-varying parameters capture COVID-19 dynamics well. • Modeling COVID-19 dynamics by coupling the compartmental model and neural networks. • Using neural networks to express time-varying parameters in compartmental models. • Applying Fourier transformation to reduce the stochastic and noise of real-world data. • Analyzing the effect of intervention policies and providing predictions.
【저자키워드】 COVID-19, Deep neural networks, Parameter estimation, Compartmental models, Runge–Kutta method, 34A34, 68T07,