Heterogeneous Geospatial Knowledge Discovery Methods for Identification of Food Desert Defection and Public Health Disparity Areas

Food deserts are geographical areas where residents face challenges in acquiring healthy and affordable food options. This lack of access, often due to both geographic and socioeconomic factors, negatively impacts public health and individual well-being. In the interdisciplinary field of urban studies, public health, and medicine, the emergent issues of food deserts contribute significantly to public health problems such as obesity, diabetes, and cardiovascular diseases. While extensive research has investigated food access disparities, there is a lack of consistent methodology for defining and identifying both geographic and socioeconomic factors that contribute to the existence of food deserts. A critical barrier in this multidisciplinary research area remains the difficulty of processing and integrating geospatial information from multiple data domains to discover comprehensive and coherent views of geospatial knowledge with the aim of identifying food deserts. This study proposes a novel geospatial knowledge discovery method involving multi-domain data integration techniques including geospatial, demographic, socio-economic, and health-related data points. After consolidating the data, ground truths are labeled using a simple ranking algorithm and machine learning models are utilized to streamline the food desert identification process. The predictive models in this study demonstrated significant potential for identifying areas at risk of becoming food deserts and for classifying the severity of associated health conditions. By carefully incorporating socioeconomic features into our models, we propose a way of predicting food desert locations with double-digit increased accuracy from the existing work and provide predictive insights into prevalence of certain diseases. Furthermore, a correlation analysis between the chosen features and food desert score is conducted to evaluate the findings of our research.