Modelling and predicting the growth dynamics of covid-19 pandemic: A comparative study including Turkey.

Estimating the growth dynamics of a pandemic is critical for policy makers to fine-tune emergency policies in health and other public sectors. The paper presents country-level calibration and prediction results on some well-known models in the literature, namely, the logistic, exponential, Gompertz, SIR and SEIR models. The models are implemented on real data from various countries, including Turkey, and their performance for different estimation windows have been analyzed using R2 scores. The computational results are obtained using Python. The Gompertz model outperforms other models by consistently offering a better tit for the total number of infected. The exponential model is helpful in describing the growth dynamics in the early stages of the COVID-19 pandemic. Suspected-Infected-Recovered (SIR) and Susceptible-Exposed-Infectious-Removed (SEIR) models display a fair performance on the underlying active cases data in many circumstances. Quantitative models can offer policy makers in Turkey and elsewhere a better insight on the evolution of pandemic when everything else is held constant and the infections follow a typical path. The results can be highly sensitive to changes in policies. There is not a single model that can perfectly mimic all stages of pandemic. An ensemble model or multi-modal distributions can be used to capture the evolution of multi-wave pandemics. [ABSTRACT FROM AUTHOR]