A Spatial Decision Support System for Evaluating POI Addition Suitability via Peer-Relative Context
Abstract
Within urban geographic information analysis, evaluating the functional composition of spatial grids is critical for understanding regional development. Existing approaches often rely on absolute density estimation or isolated supply-demand metrics, which struggle to determine if a specific region genuinely possesses the structural capacity to support new Point of Interest (POI) additions. To address this limitation, we formulate POI configuration balance as a peer-relative structural assessment. We propose a Spatial Decision Support System that evaluates the suitability of POI additions by learning the expected functional composition from surrounding urban contexts. Our framework leverages TF-IDF preprocessing to highlight functionally discriminative POIs, constructs a k-nearest-neighbor graph to capture functional peer relationships, and employs a Denoising Autoencoder optimized jointly with a Fuzzy Simplicial Set Cross-Entropy (FSCE) loss to preserve high-dimensional topology in a low-dimensional latent space. Experiments on real-world check-in data show that the learned representation retains the POI composition while preserving peer relationships, and case studies illustrate how the system can provide urban planners with an interpretable, data-driven tool to explore facility expansion scenarios.