| Abstract: |
| Infectious disease epidemics are a challenge for medical and public health practitioners. They requires prompt treatment, but it is challenging to recognize and define epidemics in real time. By knowing the short-term prediction of an infectious disease epidemic, the disease`s impact can be evaluated by preventive efforts. Real-time mathematical epidemic models such as logistic differential equations and deep learning methods are key preventative tools. Data-driven deep learning enables effective algorithms for identifying parameters in mathematical models. This paper introduces a logistic-informed neural networks algorithm inspired by applying a physics-informed neutral network to a logistic differential equation to learn the constant parameter and the time-dependent function of the Omicron variant. The learned parameter and time-dependent function, as well as the analytical solution of the logistic differential equation, are used to make a short-time prediction on the daily, the time that a plateau will be reached, and the cumulative number of individuals reported to be infected with the Omicron variant. In a data-driven simulation, the accuracy of this model is demonstrated using error metrics on Omicron variant data for Portugal, Italy, and China. |
|