Covid-19 cases continue to ebb and flow along with the dynamics of its spread and genetic mutations with new variants that continue to be studied. The quite massive impact of Covid-19 makes it important to monitor cases to be aware of new outbreaks of Covid-19. In this study, we propose a SEIR model integrated with machine learning which can provide estimation and prediction of recorded and unrecorded cases. Time- dependent infection rates are considered to accommodate diverse control and intervention processes. Generation of dynamic operators from cumulative case data is implemented to obtain the overall dynamics of the model. Furthermore, the daily unrecorded infection can be estimated based on the ratio between IFR and CFR. The rate of infection and the number of cases which are the key information in this infection are then processed using machine learning to see the characteristics of the data and estimate the rate of future infections. The simulation was carried out using infection data provided by the Jakarta City Health Office. It appears that a higher number of daily PCR tests contributes directly to a decrease in the effective reproduction ratio as well as a good case prediction test. Several simulations related to data, infection rate, and effective basic reproduction ratio are also presented. This method directly measures and predicts daily transmission indicators, which can be used effectively for epidemic control.

Covid-19, SEIR model, machine learning, dynamic generator, unrecorded cases

• Dr. Alfi Yusrotis Zakiyyah, S.Pd., M.Si.

• Felix Indra Kurniadi, S.Kom., M.Kom.

• Dr. Muhammad Fakhruddin, S.Si., M.Si.