A novel spatio-temporal prediction model of epidemic spread integrating cellular automata with agent-based modeling

Since 2019, major infectious disease outbreaks have placed tremendous pressure on global public health systems, triggering extensive research on the predictive modeling of infectious diseases. Cellular Automaton (CA) is primarily used in the spatial prediction of infectious diseases to establish a model to for simulating the interaction between different regions and the infection risk to simulate the transmission process of the disease and predict its development trend. However, CA models are governed by initial fixed rules and local interactions, and often fail to capture the complex dynamics of epidemic transmission, which are influenced by factors such as public behavior and government intervention. In view of these limitations, we propose a factorial simulation model for the spatial spread of epidemics, the CA-ABM, which divides agents into three categories-public, government, and hospital agents-to comprehensively express the macro factors that affect the development of epidemics. Agent-Based Modeling (ABM) influences the transition rules of the CA through agent choices, constraints and supporting behaviors. Focusing on the COVID-19 pandemic in mainland China from February 6 to March 20, 2020, we simulate its spread. The results showed an average improvement of 8.4 % in prediction accuracy, with few errors, RMSE under 200, and R2 values over 0.9 in most provinces, demonstrating strong macro-scale stability. This approach helps regions to understand influencing factors and enables targeted infection risk assessment and prevention. In addition, scenario analysis based on CA-ABM model changes epidemic decision-making from "prediction-response" to "scenario-response" and provides theoretical reference for future epidemic management.