.png){kind=small}
Leveraging GeoAI and Agent-Based Modeling to Address Infectious Diseases
Mr TANG Ka Chung
(Supervisor: Prof Peter K. Koh)
Abstract:
In the last five years, the COVID-19 pandemic has had a detrimental impact on both the global economy and public health, particularly in urban areas where a majority of infections have been concentrated. Despite the development of numerous predictive models by researchers worldwide, there remains a significant gap in establishing enduring, city-specific models for effective policymaking. The ongoing trend of urbanization further amplifies the risk of infectious diseases, underscoring the critical need to address this gap as a key priority in future epidemic prevention efforts. Effectively preventing epidemics requires a thorough understanding of how infectious diseases propagate, with a focus on the influence of human habitats, behaviors, and population dynamics on virus transmission. This study aims to investigate the geographical attributes of high-risk locations frequented by confirmed COVID-19 Omicron cases, analyzing the spatiotemporal diffusion patterns of the virus and the unique characteristics of its spread trajectory. Furthermore, the study will forecast epidemic dynamics under various contentious interventions such as Compulsory Universal Testing (CUT) and lockdowns.
To achieve these goals, this research initiative chose Hong Kong as a pilot location and gathered a diverse array of datasets from the Hong Kong government. These datasets encompassed census information, population movement data, air quality statistics, Point of Interest (POI) data, land usage specifics, building details, and mobility data pertaining to confirmed cases of infection, incorporating over 40 variables in total. Geospatial Artificial Intelligence (GeoAI) techniques were utilized to scrutinize the geographic attributes of high-risk zones and the characteristics of the spatiotemporal spread pattern of COVID-19 Omicron. Among the artificial intelligence methodologies employed, the Self-Organizing Map (SOM) was utilized to organize a complex dataset into a topological structure, aiding in the categorization of geographical features of high-risk areas and the spatial spread patterns within this structured framework. For the examination of spread patterns, the topological structure facilitated a clear visualization of temporal aspects, e.g., recurrent outbreaks within specific communities and the intensity of spread and outbreaks. Furthermore, a simulation-driven strategy known as Agent-Based Modeling (ABM) was executed to evaluate the efficacy of anti-epidemic measures. This ABM was integrated with an enhanced Susceptible-Exposed-Infected-Recovered (SEIR) model, the Wells-Riley model, and a mobility component to replicate the infection process and derive critical epidemic dynamics parameters, which included the effective reproduction rate, the count of exposed cases within each community, the numbers of infected, deceased, and recovered individuals. The model also can incorporate various interventions, such as mask-wearing, CUT, lockdowns, and others. By mimicking implemented anti-epidemic measures as a baseline scenario, this study compared the effectiveness of additional interventions like CUT and lockdowns. 3
The results of this research offer valuable recommendations and strategies for both short-term and long-term prevention efforts in the future. Long-term strategies encompass adjustments in land usage, strategic placement of public amenities and retail establishments, and enhancements in air quality. Short-term measures involve implementing interventions during pandemics and prioritizing community-wide infection testing. These insights are anticipated to support policymakers effectively in averting infectious disease outbreaks, thereby reducing infection rates and mortality rates.
Comments