연관규칙 마이닝과 마코프 전이 분석을 활용한 사업관리단계 BIM 설계이슈의 발생 메커니즘 탐색

Abstract

Building Information Modeling supports multidisciplinary collaboration, yet recurring issues such as clashes, omissions, and discrepancies often propagate across domains, amplifying project risks. Existing approaches mainly detect isolated errors and overlook interdependent and cascading patterns. This study integrates association rule mining and Markov chain analysis to reveal both static co-occurrence and dynamic transition pathways of BIM issues. Using pre-construction issue logs, 39 significant rules were identified, including intra-domain accumulations (e.g., drawing discrepancy leading to information omission) and cross-domain propagations (e.g., structural information omission triggering architectural drawing discrepancy). Markov modeling further highlighted recurrent loops among omissions, discrepancies, and review requests, as well as pathways from mechanical clashes to architectural discrepancies. Findings provide predictive insights, preventive strategies for blocking repetitive loops, and tailored management through metadata standardization and checklist-based safeguards. The study demonstrates the value of combining data-mining and probabilistic modeling for proactive BIM quality management, while noting limitations in dataset scope and temporal representation.