Comparative Analysis of Urban Functional Zone Identification Methods Based on POI Data: Improvements Using Network Constraints and Cell Declustering

Urban functional zones (UFZs) are essential for understanding spatial structure and optimizing residents' welfare. While Point of Interest (POI) data serve as a primary source for identifying UFZs, existing studies often overlook data bias, such as spatial clustering, and lack systematic comparison of identification methods. To address these gaps, this study develops a comparative framework for mining UFZ patterns in the central urban area of Wuhan. We compare the Frequency Density (FD) method against Kernel Density Estimation (KDE) and Network‐constrained Kernel Density Estimation (NcKDE) and innovatively introduce a cell declustering weight adjustment method to mitigate bias caused by POI clustering. Results are validated qualitatively and quantitatively using official plans and remote sensing imagery. The comparative analysis reveals that (1) the NcKDE method achieves the highest overall accuracy (0.97), demonstrating its effectiveness for identifying mixed‐functional areas by accounting for road network constraints and (2) the cell declustering method significantly improves the baseline FD accuracy from 0.77 to 0.87, making it the optimal approach for identifying single‐functional zones. From a spatial governance perspective, this study analyzes the discrepancies between the identified actual urban functions and official top‐down planning. These insights provide scientific support for coordinating spatial functions, enhancing living quality, and promoting people‐centered urban development.