Association Rules of BIM Clash Detection Data using Data Mining and Attribute Sensitivity Analysis

The BIM design conversion phase often suffers from recurring coordination errors due to multi-disciplinary model integration. This study applies association rule mining to real BIM issue data to extract statistically significant patterns, validated through Fisher’s Exact Test. Frequent patterns include repeated omissions in structural/architectural models at lower floors, cross-disci-plinary error propagation from mechanical clashes to architectural discrepancies, and intra-model transitions of drawing-review issues into omissions or discrepancies. These patterns highlight the need for consistent coordinate systems, reference standards, and rigorous issue logging. Subsequently, sensitivity analysis of four attributes—trade, floor, issue type, severity—revealed that floor is critical for rule precision, while trade enhances pattern diversity and accountability. Issue type and severity are auxiliary, contributing to classification detail and response prioritization. Findings suggest managing floor and trade as mandatory standardized metadata, supported by validation constraints and checklist-based reviews, to enable proactive detection and prevention of design conversion errors.