This study aims to identify the regions and people with low food access (LFA) for Korea at the national level and to examine disparities in food consumption, dietary behavior, and health outcome for those regions and people. Based on the distance to the nearest grocery store from residence, the regions and people with LFA are identified through geographical information system (GIS) analysis. To examine disparities between the regions and people with LFA and without LFA, a consumer survey is conducted and data from National Health and Nutrition Examination Survey and Community Health Survey are analyzed. This study found that there exists a serious access to food issue in Korea, especially for the aged. Moreover, there also exist significant disparities between the regions and people with and without LFA in the distance and one-way travel time to the grocery store that is mainly visited, frequency of offline/online grocery shopping, availability of various foods, dietary habits such as eating regularly, eating nutritionally balanced foods, and eating sufficient fruit/vegetable/whole grains, the acquisition and utilization of food-related information, and health outcomes. This study suggests that, to resolve such a serious food access problem, assistance policies, such as mobile grocery stores and lunch-box delivery, need to be activated in countries similar to Korea since this problem could potentially deteriorate the national medical finances as well as the regional and individual disparities.

The small commercial stores opening in housing structures in Seoul have been soaring since the beginning of this century. While commercialization generally increases urban vitality and achieves land use mix, cafés and restaurants in low-rise residential areas may attract numerous passenger populations, with increased noise and crimes, in the residential area. The urban commercialization is so fast and prevalent that neither urban researchers nor policymakers can respond to it timely without a practical prediction tool. Focusing on cafés and restaurants, we propose an XGBoost machine learning model that can predict commercial store openings in urban residential areas and further play the role of an early warning system. Our findings highlight a large degree of difference in the predictor importance between the variables used in our machine learning model. The most important predictor relates to land price, indicating that economic motivation leads to the conversion of urban housing to small cafés and restaurants. The Mapo neighborhood is predicted to be the most prone to the commercialization of urban housing, therefore, its urgency to be prepared against expected commercialization deserves underscoring. Overall, our results show that the machine learning approach can be applied to predict changes in land uses and contribute to timely policy designs in rapidly changing urban context.