![This photo, taken on July 23, 2020, shows health workers guiding arrivals at Incheon International Airport in Incheon. (Yonhap)](http://koreabizwire.com/wp/wp-content/uploads/2020/07/PYH2020072306490001300.jpg)
This photo, taken on July 23, 2020, shows health workers guiding arrivals at Incheon International Airport in Incheon. (Yonhap)
SEOUL, Aug. 20 (Korea Bizwire) — A South Korean research team has developed a new artificial intelligence (AI)-based model that can predict the number of foreign COVID-19 cases entering the country.
Researchers from the Korea Advanced Institute of Science and Technology announced on Wednesday that it had developed a new AI model that was capable of predicting the number of foreigners infected with COVID-19 entering South Korea.
By using data such as the number of COVID-19 patients and deaths in foreign countries and the number of daily flights to Korea, the research team developed a model to predict the number of foreigners with COVID-19 entering the country over the following two weeks.
To calculate the COVID-19 risk of foreign countries, the research team used the frequency of COVID-19 related keyword searches, in addition to the reported number of COVID-19 patients and deaths.
Geological correlation and the degree of exchanges were also included as key factors.
The research team dubbed the research model ‘Hi-COVIDNet.”
On the basis of the data learned during the period from March 22 to May 5, the research team predicted the number of foreigners infected with COVID-19 entering the country over the following two weeks (May 6 to 19).
The prediction had a margin of error of just two to three persons compared to the actual number of incoming foreign COVID-19 patients during the period.
The research team explained that the accuracy of the newly-developed AI model was about 35 percent higher on average than that of time series data-based machine learning deep learning prediction models, which have been used primarily for the analysis of stock prices.
J. S. Shin (js_shin@koreabizwire.com)