시계열 모형과 빅데이터 분석기법을 이용한 코로나 확진자 수 예측

Abstract

Purpose: In this study, we used time series models, deep learning (DL) models, ensemble models, and other models to predict COVID-19 confirmed cases. We developed time series models with exogenous variables and achieved show promising results for the correlation between COVID-19 confirmed cases and vaccination rates. Methods: We proposed a method based on time series and deep learning for model development. The proposed method can accurately predict the number of confirmed cases of COVID-19 per day by utilizing the COVID-19 vaccination rate as an exogenous variable. Thus, improved prediction accuracy can be achieved using DL ensemble models. Results: The AdaBoost-LSTM model yielded superior results than the other time series models, and the SARIMAX(3rd vaccination rate)(2,1,3)(1,1,1,7) model exhibited better prediction performance than other time series models. Conclusion: The SARIMAX(2,1,3)(1,1,1,7) model exhibited better performance than gated recurrent unit/long short-term memory models. The use of the AdaBoost algorithm improved the prediction performance of the model by approximately 51.6%.