Prediction of epidemic parameters using machine learning methods

Epidemics have occurred at different times throughout history for various reasons. Epidemics can cause various consequences such as loss of life, deterioration of the social and economic structure. Therefore, countries take some measures to reduce the impact of the epidemic. Epidemic modelling methods and optimization algorithms are used to evaluate the potential costs and benefits of the measures taken in the fight against the epidemic. In this thesis, studies were carried out to model the COVID-19 epidemic, which has affected the world today, and to predict its parameters. Data from the COVID-19 outbreak in Egypt and South Korea were used in the experiments. First, optimizations were carried out using the L-BFGS-B, Powell, CG, COBYLA, and SLSQP methods on the SEIR model, which is widely used as an epidemic modelling method. Then, predictions were made using calculated values such as the R0 value, mortality rate, and transition rates between epidemic parameters. Using the same datasets, predictions were also made using the machine learning methods LSTM and DVM. The prediction successes of the results obtained from SEIR, LSTM, and DVM methods were compared using the MAPE metric, and the results were evaluated with graphics. The MAPE value of 79% of the 24 different prediction results was below 10