Clash Relevance Prediction in BIM-Based Design Coordination Using Bayesian Statistics

Multiple disciplines in construction projects have greatly improved the efficiency of their coordination efforts by using building information modeling (BIM) for clash detection. However, because the outcome of clash detection includes many irrelevant clashes that have no substantial influence on a project, the precision of the method has been questioned. To address this problem, this paper uses Bayesian statistics to distinguish relevant from irrelevant clashes for improving the clash detection of BIM. The paper compares naive Bayesian, the Bayesian network, and Bayesian probit regression, and validates the effectiveness of each method. Additionally, the paper discusses how prediction can be improved by combining the three methods by majority rule. Bayesian statistics provide a method of mining knowledge from historical data and leads to clash management processes that are more independent of the project experience of BIM coordinators.