공공데이터 주소 품질 향상을 위한 절차적 정제 프레임워크(AQRF): 설계와 실증

초록

This study proposes the Address Quality Refinement Framework (AQRF), a process-based approach to systematically improve the quality of address data in public datasets. AQRF comprises four stages—Semantic Decomposition, Component Completion, Administrative Validation, and Error Correction—and manages address quality across the data lifecycle through an iterative loop of refinement, validation, and evaluation. We conduct an empirical assessment on approximately 14,000 records drawn from five healthcare-related public datasets (medical institutions, pharmacies, retail/clinic listings, health information exchange participants, and workers’compensation medical institutions). The average refinement success rate reached 99.18%, and both component recognition and administrative consistency—measured with reference to KS X ISO 19157—showed consistent improvements. Error analysis indicates that component recognition failures were the most prevalent, largely attributable to the coexistence of parcel-based and road-name address systems and heterogeneous data entry practices across agencies. The principal contribution is to reconceptualize address quality as a structural relationship across datasets rather than a mere field-level accuracy issue, extending the unit of quality management to the process itself. AQRF serves as a reference model that enhances data linkage, interoperability, and AI-readiness, and it aligns with administrative codes (legal/administrative district, road-name codes) and international quality standards, thereby supporting the institutionalization and standardization of public-sector data governance.

키워드