Options
საქართველოში კორონავირუსის გავრცელების პროგნოზირება
Date Issued
2020
Author(s)
Advisor(s)
Institution
Abstract
Since the first outbreak of the epidemic in China in December 2019, various models have been
developed and used to predict coronavirus. The paper focuses on predicting the spread of
COVID 19 and evaluating preventive measures taken by the government. The study was
conducted by employing Georgian coronavirus data. Epidemiological models, classical
econometric time series models, and also machine learning models were used for forecasting.
Furthermore, considering the nature and trend of the spread of the epidemic, we have proposed
a polynomial model with modified architecture. The paper also discusses several
parameters/ratios of the epidemic. The most important one, the number of reproductions, was
modeled with several approaches. The result was not identical but comparable to certain
circumstances.
The results of the study showed that it is difficult to construct an ideal model for prediction.
Each model has its advantages. In particular, phenomenological models well predict the
epidemic trend, duration, and the total number of infected people, although they make a big
mistake when predicting daily cases. In contrast, machine learning models predict relatively
accurate daily infection in the short term, although due to data limitations, it is difficult to
predict in the long run. Furthermore, compartmental models are the best choice for modeling
state-controlled restrictive measures and determining the optimal level of restraint.
developed and used to predict coronavirus. The paper focuses on predicting the spread of
COVID 19 and evaluating preventive measures taken by the government. The study was
conducted by employing Georgian coronavirus data. Epidemiological models, classical
econometric time series models, and also machine learning models were used for forecasting.
Furthermore, considering the nature and trend of the spread of the epidemic, we have proposed
a polynomial model with modified architecture. The paper also discusses several
parameters/ratios of the epidemic. The most important one, the number of reproductions, was
modeled with several approaches. The result was not identical but comparable to certain
circumstances.
The results of the study showed that it is difficult to construct an ideal model for prediction.
Each model has its advantages. In particular, phenomenological models well predict the
epidemic trend, duration, and the total number of infected people, although they make a big
mistake when predicting daily cases. In contrast, machine learning models predict relatively
accurate daily infection in the short term, although due to data limitations, it is difficult to
predict in the long run. Furthermore, compartmental models are the best choice for modeling
state-controlled restrictive measures and determining the optimal level of restraint.
File(s)
Loading...
Name
Samagistro Gaprindashvili.pdf
Description
საქართველოში კორონავირუსის გავრცელების პროგნოზირება (ეკონომეტრიკული და მანქანური სწავლების მეთოდები)
Size
1.93 MB
Format
Adobe PDF
Checksum
(MD5):698a0bf735c4a393cf98baaa41f28ef4