RNN enhanced Compartmental Model for Infectious Disease Prediction

COVID-19 has caused a pandemic on a global scale and brought about harm to politics, economy, and other fields. It is of great significance to build an epidemic analysis model for better understanding pandemic transmission patterns and formulating anti-epidemic policies. However, the performance of previous epidemic analysis methods is limited by the parameter estimation problems and the lack of interpretability. In this paper, we proposed RNN enhanced compartmental dynamic model with time-varying parameters for COVID-19 epidemic analysis, called RE-SEIDR model. This model can utilize RNNs to capture the transmission pattern varies caused by anti-epidemic policies changes, and the mutation of the virus, and so on, and then predict time-varying parameters of the compartmental model. After that, the problem of inconsistency between compartmental models’ time-invariant parameters and real COVID-19 transmission patterns is resolved. By conducting experiments across epidemic data from six European countries, we demonstrated that our model outperforms existing state-of-the-art baselines in terms of the active confirmed cases prediction. We also demonstrated that our model can provide an in-depth understanding of the transmission mechanism of the pandemic by the effective reproductive number analysis and time-varying parameters analysis, which may help public health res develop more effective antiepidemic policies.