(368b) A Conceptual Study on the Epidemic Spread Based on Cellular Automata and Feature Extraction

With the progress in process systematic theory and modelling practice, more and more social and biological system can be studied as a generalized process system, the spread of infectious disease can fall into the category of such systems, which involve emergence of initial infection, diffusion-like propagation across the geographic area and time, and have been mathematically studied since the beginning of last century. Amongst various mathematical methods for describing the spread of infectious diseases, cellular automata make it feasible to explicitly simulate the spatial and temporal evaluation of the epidemic with intuitive local rules obtained from the mechanism of spread for a specific disease. Unfortunately, these local rules are usually covered by disturbed symptoms and suffering of patients, and are available to public through a long-term epidemiological observation and analysis. The development of data mining and machine learning techniques significantly facilitate the feature extraction in this area.

In this paper, a multi-scale model is proposed and realized on a cellular automata platform. This model considers a simplified social system with various parameters, including infection rates, infection and quarantine period, population movement, as well as the citizen awareness. When all these factors exist at the same time and interact over time in the system, the critical factors for epidemic control become diverse and unpredictable, which makes the situation even more challenge for healthcare professionals and decision makers. With proper data-driven feature extraction method, the critical factors for epidemic spread can be identified for further situation control.

All the impact factors mentioned above are tested by adjusting the corresponding parameters in this model, so that the quantitative influence of those factors are investigated respectively. Based on our results, an average incubation period hidden in the model mechanism can be identified by data analysis, which is key information for the development of quarantine strategies, and the intergenerational evolution of virus can be obtained through infectivity decay to understand the current status of epidemics and obtain a reference for adjusting the clinic treatment and control strategy accordingly.