Outbreak detection of coronavirus disease epidemic: Transmission dynamics using machine learning techniques (Preprint)

Abstract

In the face of the rapid spread of the coronavirus disease (COVID-19) pandemic, predicting the spread of COVID-19 can help healthcare providers prepare and respond to outbreaks more quickly and effectively, ultimately leading to better care for patients.

We aimed to develop a method to detect early COVID-19 outbreaks or identify potential early outbreaks using machine learning (ML) by analyzing epidemiological data in the Republic of Korea.

ML methods were developed to predict the transmission trend of COVID-19, and a new method was proposed for detecting the start time of new outbreaks. We constructed risk index and label to measure the change in the transmission trend of COVID-19. Additionally, the study used ML methods to detect the start time of new outbreaks based on Label 2, which was maintained for at least 14 days, and predicted labels to detect future transmission trends early.

ML methods had a high accuracy of more than 90% in estimating the classification of the transmission trend (Increase, Maintain, or Decrease). The proposed method not only accurately predicted the start time of new major outbreaks but also provided a standard for accurately predicting the start time of minor outbreaks. Random Forest exhibited the most accurate estimation of outbreak detection.

The proposed method can provide an explainable standard for accurately predicting the start time of both major and minor outbreaks. This can help develop effective prevention and control measures for COVID-19 transmission.