BIM log mining: exploring design productivity characteristics

Abstract

A clustering-based building information modeling (BIM) log mining method is developed in this research to provide a data-driven knowledge discovery about the design productivity characteristics from a huge amount of BIM design log data. Since design behaviors are non-deterministic and subjective, a novel clustering algorithm named efficient fuzzy Kohonen clustering network (EFKCN) is utilized to produce informative clusters of different features. First, datasets are pulled out from the raw design logs and transformed into understandable forms for computers. Then, EFKCN clustering algorithm is performed in datasets at both individual and team level. Finally, analysis and prediction methods, like time analysis, regression and others, will further investigate the extracted clusters, which help managers to figure out the design preference and productivity of different designers. A case study is conducted in the real BIM design logs from an international architecture design firm with 853,520 records to illustrate the effectiveness of the proposed method. From a view of individuals, the personal design behaviors hidden in different clusters are served to arrange proper design work rationally during particular time periods. From a team perspective, the design productivity of different designers can be approximately evaluated as high, medium, and low levels in an objective and efficient manner. In the comparison of EFKCN with other popular clustering methods, EFKCN takes less time and iterations to complete the clustering process, and its clustering results achieve great compactness and separation according to the value of cluster validity indices (CVIs). Hence, this research contributes to performing the novel clustering-based BIM log mining, which acts as a powerful decision-making tool in evaluating design productivity and drawing up personalized work arrangements for a more efficient modeling process.