Prediction of The Number of COVID-19 Cases Using LSTM, ARIMA, and SARIMAX Models
Presenter Type
UNO Graduate Student (Masters)
Major/Field of Study
Mathematics
Other
Data Science
Author ORCID Identifier
0000-0002-0748-156X
Advisor Information
Xiaoyue Zoe Cheng
Location
MBSC Ballroom Poster # 608 - G (Masters)
Presentation Type
Poster
Start Date
24-3-2023 2:30 PM
End Date
24-3-2023 3:45 PM
Abstract
The extremely contagious COVID-19 virus was discovered in Wuhan, China, in December 2019. Governments and decision-makers were battling to restrict COVID-19's dissemination because it emerged as one of the most pressing global issues. They used highly tight Non-Pharmaceutical Interventions (NPIs) in the early stages of the COVID-19 pandemic to bring it under control because of its rapid spread and the lack of an effective treatment or vaccine. Recent research shows that those extreme cases of NPIs have played a crucial role in controlling the spread of not only COVID-19 but also a variety of other viruses and diseases. The main objective of this study is 2 weeks prediction of the number of COVID-19 cases in the US by performing Long Short-Term Memory (LSTM), ARIMA, and SARIMAX models, feeding 2020 historical (NPIs), individual temperature records from smart thermometers dispersed across the US (> 1 million users), and the specific tweet phrase "fever" as the most common symptom among infected cases. This study's findings show that the ARIMA model performs best at forecasting the daily number of infected cases using 100 days' worth of historical infected cases. In comparison to the SARIMAX model, LSTM performs better when exogenous variables are taken into account.
Scheduling
2:30 -3:45 p.m.
Prediction of The Number of COVID-19 Cases Using LSTM, ARIMA, and SARIMAX Models
MBSC Ballroom Poster # 608 - G (Masters)
The extremely contagious COVID-19 virus was discovered in Wuhan, China, in December 2019. Governments and decision-makers were battling to restrict COVID-19's dissemination because it emerged as one of the most pressing global issues. They used highly tight Non-Pharmaceutical Interventions (NPIs) in the early stages of the COVID-19 pandemic to bring it under control because of its rapid spread and the lack of an effective treatment or vaccine. Recent research shows that those extreme cases of NPIs have played a crucial role in controlling the spread of not only COVID-19 but also a variety of other viruses and diseases. The main objective of this study is 2 weeks prediction of the number of COVID-19 cases in the US by performing Long Short-Term Memory (LSTM), ARIMA, and SARIMAX models, feeding 2020 historical (NPIs), individual temperature records from smart thermometers dispersed across the US (> 1 million users), and the specific tweet phrase "fever" as the most common symptom among infected cases. This study's findings show that the ARIMA model performs best at forecasting the daily number of infected cases using 100 days' worth of historical infected cases. In comparison to the SARIMAX model, LSTM performs better when exogenous variables are taken into account.