Prediction of Disease Outbreaks with the SIR Model and Richards Model in Multi-wave Cases

Modeling the spread of infectious diseases is mainly done to overcome the COVID-19 pandemic until mid-2022. The models used include analytical, stochastic, and phenomenological. In this paper, the authors predict the case of multi-wave by comparing two models, namely the phenomenological model represented by the Richards curve model and the stochastic model represented by the multi-state SIR model. This study aimed to obtain the best model for short-term predictions of COVID-19 cases. Since COVID-19 cases occur in multi-wave, we added a changepoint detection method to determine the wave boundary. Changepoint detection helps to get the last wave boundary which is used as the basis for data collection for modeling using the multi-state SIR model and the Richards curve model. In this article, we adjust daily cumulative epidemiological data for COVID-19 using a multi-wave model with a Richards curve. To overcome the multi-wave pandemic, we carried out changepoint detection using the binary segmentation method. The analysis shows that the prediction results using the multi-state SIR model with the assumption of discrete-time Markov chain are more accurate than the Richards curve model. However, the difference in the MAPE values obtained is not significant. The conclusion is that the prediction results with the multi-state SIR model will be more accurate if it uses more states according to the case that occurs. Thus the prediction results using the Richards curve are valid with simple steps.