Prediction of COVID-19 Death Rates with Artificial Neural Network Under Different Time Series Models Based on Moving Averages

Çaparoğlu Ö. F., Ok Y., Tutam M.

1 st International Conference on Advances in Engineering, Architecture, Science and Technology, Erzurum, Türkiye, 15 - 17 Aralık 2021, ss.303-309

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Erzurum
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.303-309
  • Atatürk Üniversitesi Adresli: Evet

Özet

The novel corona virus disease (Covid-19), spreading more than 230 million people and causing nearly 5 million deaths as of October 2021, was declared as a global pandemic by the World Health Organization in March 2020. Accurate and consistent estimation of incidences is utmost importance for countries to combat the Covid-19 pandemic and its consequences. One of the most important indicators showing the effective fight against the Covid19 is the number of deaths. In this study, different Artificial Neural Network designs and time series models are evaluated, and the best network design is suggested to predict the future death numbers with a high accuracy rate. We find that the best estimation of death numbers relies on historical data with 11 days of moving average model for the number of tests and cases, as well as 10-13 days of weighted moving average model for the number of seriously ill patients.

Keywords. Covid-19, Prediction, Time-Series, ANN, Death Numbers