Machine Learning Based 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 present complex challenges, particularly in urban contexts where they contribute significantly to public health problems such as obesity, diabetes, and cardiovascular diseases. While numerous 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 diverse data domains in order 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 with multi-domain data integration techniques for diverse datasets including geospatial, demographic, socio-economic, and health-related ones. 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 incorporating disease prevalence rates and socio-economic features into our models, we not only propose a way of predicting food desert locations with 11% increased accuracy from the existing work but also provide predictive insights into public health outcomes.