Using GeoAI and public data to control and manage infectious diseases in a highly dense city—an application of Self-Organizing Map approach to COVID-19 in Hong Kong

Keywords: COVID-19,Self-Organizing Map,Sociodemographic characteristics,Hong Kong
Abstract Type: Paper Abstract

Abstract

Background: During the COVID-19 pandemic, public health authorities across the globe have tried to trace infectious patients. However, detecting new, highly contagious viruses efficiently and effectively is challenging. This study, therefore, aims to present a novel approach to examine the spatiotemporal associations of the COVID-19 spread patterns and sociodemographic characteristics in Hong Kong using Self-Organizing Map (SOM) and the existing public data.

Methodology: A SOM algorithm, a class of AI techniques, was applied to classify the clusters of districts with similar levels of COVID-19 transmission and sociodemographic characteristics using trained neurons. Compared to other clustering algorithms, SOM is easier to visualize the distance among clusters by calculating Best Matching Unit (BMU). We used the daily reports of the building list with COVID-19 cases from March 2020 (i.e., the second wave) to May 2021 (i.e., the fourth wave) and the 2021 population census in Hong Kong.

Result: During the second wave, high-income districts and commercial areas were likely to be high-risk areas. Low-income districts with high population density and high working populations were identified high-risk areas during the third wave. In addition, the transmission gradually shifted from east to west of Hong Kong. During the fourth wave (with the Alpha and Delta variants), the transmission spread from low-income districts to high-income districts.

Conclusion: SOM can effectively identify and predict potential high-risk areas of infectious disease transmission based on the characteristics of the community’s sociodemographic and the diseases. Therefore, the use of SOM can be pivotal for future disease control and management.