Parameter Estimation for a Modified SEIR Model of the COVID-19 Dynamics in the Philippines using Genetic Algorithm

Abstract

The Philippines has been under a series of different levels of community quarantine and this affected the dynamics of the COVID-19 spread in the country. Predicting the trajectory has been an interest of various research groups. To provide a more efficient method to estimate the parameters of the Age-Stratified, Quarantine-modified SEIR model with Nonlinear Incidence Rates (ASQ-SEIR-NLIR) other than the shooting method, a genetic algorithm approach is explored. By defining constraints for each parameter, the algorithm arrived at an acceptable optimal value for each parameter. The experiment is done on two regions of interest: the Philippines (country-level) and Quezon City, Metro Manila (city-level). The ASQ-SEIR-NLIR model, using the parameters generated by the genetic algorithm, is able to produce an average trajectory compared to the actual data, which may be deemed noisy. The dynamics of the COVID-19 spread between Quezon City and average country level is compared, showing that the city population is being exposed to the virus at a much faster rate than the country average and may have more asymptomatics not getting tested than the country average. Given the average trajectory, the peak daily infection projection is way lower at 0.0823% of the country population for the country projection and 0.1494% of the Quezon City population for the city projection, which is below than previous literature estimates of 3-10%.