Exploring the accuracy of an optimization-free neural network forecasting model in mathematical epidemiology: A case study in Turkiye

In this study, we explore the use of mathematical epidemiology models in predicting COVID-19 cases in Turkey. Our approach employs a Feed-Forward Neural Network solver, which is designed to quickly converge and make accurate predictions. To eliminate the need for time-intensive optimization procedures, the network weights are calculated using the Extreme Learning Machine algorithm, ensuring adherence to the initial conditions set by the epidemiology models. We examine the performance of both the Susceptible-Infected (SI) and Susceptible-Infected-Susceptible (SIS) models using this approach and evaluate their accuracy.