클레이독스 기반 환경 정도검사 표준 시험성적서 자동 분석 시스템 개발

초록

This study developed ATRAS (Automated Test Report Analysis System), an automated system for analyzing standardized environmental proficiency test reports extracted from the Claydox platform. Raw data exported from Claydox in CSV or XLSX formats contain inconsistencies in item names, institution names, addresses, measurement units, and judgment labels, as well as missing and erroneous entries, rendering them unsuitable for direct statistical analysis and frequently causing errors during data integration and indicator computation. To address these challenges, ATRAS was designed as an end-to-end pipeline comprising data upload, preprocessing, analysis, visualization, and result delivery. During preprocessing, column names and field structures were standardized; institution, manufacturer, and maintenance provider names were normalized using unified criteria; address and regional expressions were mapped to administrative units; inconsistencies in measurement ranges and units were corrected; and judgment labels were consolidated into binary outcomes (pass/fail). Through this process, heterogeneous test report data were transformed into analysis-ready structured datasets. In the analysis stage, ATRAS computed nonconformity rates and activity indicators across multiple dimensions, including institutions, manufacturers, regions, equipment age, and inspectors. Rule-based detection was applied to identify localized abnormal patterns, generating summarized insights to support operational decision-making. In the visualization stage, multidimensional graphical components—such as treemaps, donut and bar charts, time-series trends, and quadrant analyses—were employed to facilitate intuitive exploration of distributional concentration, temporal variation, and relationships between activity levels and quality performance. As a result, the average nonconformity rate for 980 water quality test reports was 2.96%, indicating an overall stable quality level, although localized increases were observed for equipment aged 1–2 years and 6–7 years. For 1,335 drinking water test reports, the average nonconformity rate was lower at 1.4%, while quality degradation was identified for equipment aged 7–8 years and among small-scale manufacturers. Monthly trend analysis further revealed recurring periods of elevated risk, suggesting that preventive inspections concentrated around May and September represent a reasonable operational strategy. By supporting filtering, drill-down analysis, and CSV export through Streamlit- and React-based dashboards, ATRAS demonstrates the feasibility of transforming test report–based quality management into a reproducible, efficient, and data-driven operational framework.

키워드