A POI-constrained multi-source online geocoding optimization method

Given the lack of publicly available official geocoding resources in many countries, online geocoding services are often the only option for non-specialist users to conduct spatial analyses. However, their results often exhibit highly variable positional accuracy, with frequent occurrences of large errors. This is predominantly because of their limited and outdated reference dataset, and the uniform address-matching methods employed by individual platforms. To address these issues, we propose a geocoding optimization method (GOM) that leverages multiple online geocoding platforms and updated Points of Interest (POI) data to generate POI-constrained geocoding outputs. This approach enables more accurate results by reducing the dependence on a single platform or using limited reference datasets. Using 1769 address records from the Gulou District, Nanjing, we demonstrated that the GOM outperforms all major online geocoding platforms in reducing mean positional errors and the frequency of large errors. This method improves geocoding accuracy, reduces spatial distortions and minimizes their impact on spatial statistical results and analytical outcomes, and offers a practical solution for users who require high-precision geocoded data for research and decision making.