A research group at Nagoya University and the Institute of Physical and Chemical Research has begun predicting infection with the new coronavirus infection (COVID-19) using a new method that optimally links mathematical models and measured data using a mechanism similar to weather forecasting.Utilizing the latest data obtained every day, it is possible to predict virus infection with high accuracy.
The technology applied to this method is data assimilation, which is the key to weather forecasting.By linking simulation (mathematical model) and real world (actual measurement data) based on statistical mathematics and dynamical system theory, it contributes to improvement of prediction accuracy such as mitigating the influence of noise contained in actual measurement data.
For the actual measurement data, we used three data obtained every day ((19) number of people requiring hospitalization, (XNUMX) number of people who were discharged or canceled, and (XNUMX) number of deaths).For the mathematical model, we constructed an "extended SIR model" that is an original extension of the SIR model, which is known as a typical infectious disease mathematical model, according to the characteristics of COVID-XNUMX.The advanced data assimilation method (Ensemble Kalman filter) used in weather forecasts was applied to these and fused.
From the above data assimilation, we first estimated the "effective reproduction number", which indicates how many people were infected by one infected person.Effective re
Although the number of production cannot be known directly, it is useful as an indicator of the spread of infection, and when the transition is estimated using data from all over the country, a decrease is seen during the period of the state of emergency issued to Tokyo three times in the past. It was possible to confirm the infection suppressing effect.
In addition, we released the results of future prediction of infection transition in three prediction scenarios corresponding to these infection control effects and a prediction scenario when the infection control effect is zero.In the future, it is expected to contribute to the prevention of the spread of infection and the formulation of a response plan to take the initiative by incorporating factors such as the effects of vaccination, human flow and temperature, and aiming for more precise predictions.
