Prediction and layout optimization of older adult(s) care facilities based on POI data, machine learning, and space...

As the population continues to age, the scientific selection and rational layout of older adult(s) care facilities are increasingly becoming key issues in urban planning. Yet, traditional planning methods have clear limitations in terms of spatial precision and practicality. This study focuses on Shanghai, a megacity experiencing rapid population aging. It explores a three-tiered, progressive framework for predicting the location of older adult(s) care facilities that integrates POI data, machine learning algorithms, and space syntax.

Shanghai was divided into 33,010 500 × 500-m grid cells, and a training dataset was constructed using POI data. The C5.0 decision tree algorithm was employed, and a multi-algorithm benchmark comparison was conducted with logistic regression and random forests. Through a three-tiered progressive screening process-comprising citywide grid prediction, secondary screening based on aging data, and space syntax-based local choice verification-the study sequentially identifies preliminary suitable grids, priority deployment grids, and key planning grids.

The study found that: (1) The C5.0 decision tree model achieved a prediction accuracy of 83.36%, outperforming logistic regression (82.05%) and random forests (80.96%), indicating that the method is robust and reliable. The model identified 940 preliminarily suitable grid cells, with a spatial match rate of 89% compared to existing older adult(s) care facilities (χ² = 30,517, p < 0.001). Catering, real estate, and company facilities were identified as key influencing factors, indicating that existing older adult(s) care facilities tend to cluster in areas with convenient daily services, high residential density, and accessible employment opportunities. (2) Based on the aging rates and older adult population sizes of each subdistrict, a secondary screening was conducted to identify 519 priority deployment grids. These areas exhibit a center-to-periphery decreasing distribution, with the highest density in the central urban area and few in outer suburban areas such as Chongming. (3) A citywide space syntax analysis was conducted using the complete Shanghai road network, identifying 104 key planning grids from among 519 priority deployment grids. Taking Jiading District-which has the highest number of priority deployment grids (168)-as a case study, the space syntax analysis was used to narrow the selection down to 34 key planning grids.

The three-tiered, progressive site selection framework developed in this study integrates macro-level predictions of functional suitability with micro-level assessments of road network accessibility. It provides urban planning authorities with actionable spatial decision-making references for identifying priority deployment areas for older adult(s) care facilities, and offers a replicable methodological approach for the planning of public service facilities in other high-density cities.