Abstract:

The non-Markov and Markov spreading models are two common infectious disease spreading models. The essential difference between them lies in whether the spreading process depends on the historical dynamic state. Although some studies have revealed that the two can be equivalent under certain conditions, the accuracy of this prediction in an actual epidemic is insufficient. Thus, based on the SHIJR(susceptible, hidden, infected, confirmed, removed) model, this paper compares and analyzes the predictive ability of the non-Markov and Markov spreading models for the COVID-19(coronavirus disease 2019) epidemic situation. Assuming that the system is in a uniform mixing mode, the time distribution of the state transition in the non-Markov model can be converted to the corresponding transition rate in the Markov model. Then, the optimal propagation parameters of the two modes can be simulated respectively. In the simulations that we performed, we found that the parameters simulated by the non-Markov model were more realistic, and both the short-term and long-term prediction effects were better than those for the Markov model. This work fills the gap for predictive comparisons between these two communication models and promotes a greater understanding of their predictive abilities and applicable conditions.

Key words: epidemic spreading, non-Markov property, spreading prediction model, epidemic situation