BIM log mining: Discovering social networks

This research develops a systematic methodology to deeply mine tremendous volumes of design logs (that are generated from Building Information Model (BIM) design process) to discover social networks in BIM-based collaborative design practices and examine the relationship between the characteristics of the design social network and production performance of designers. Firstly, a data extraction procedure consisting of data harvesting, parsing, and cleaning is proposed to obtain BIM design logs in a Comma Separated Values (CSV) format from several designers over the course of working on multiple projects. Secondly, a metric of working together on joint cases is proposed to build up a weighted sociogram that is consisting of performers P, relations R, and weights W. Lastly, a number of indicators are defined to measure and analyze structural characteristics of the discovered BIM-based collaborative network at macro-, meso-, and micro- levels. A dataset of design logs that involves 51 designers working on 82 projects with 620,492 lines of commands, provided by a major international design firm, is used as a case study to demonstrate the feasibility and applicability of the developed approach in this research. Results indicate that: (i) Strong positive correlations exist across all centrality measures calculated based on the discovered social network of BIM-based collaborative design where designers located in the center of the interaction map (with the greatest degree centralities), such as designers "#2" and "#24", are generally those who provide the shortest communication channels (with highest betweenness centralities) and are most reachable for others (with highest closeness centralities); and (ii) All the node centrality measures are significantly and positively related to the production performance of designers in the BIM-based collaborative network. Particularly, the measured node degree centrality by weight is capable of explaining the greatest percentage of variations (71.13%) in the production performance of designers. This research contributes to: (a) The state of knowledge by proposing a novel methodology that is capable of capturing and modeling collaborations among designers from tremendous event logs to discover social networks; and (b) The state of practice by providing insight into a better understanding of relationships between sociological network characteristics and production performance of designers within a design firm.