A data mining trajectory clustering methodology for modeling indoor design space utilization

Traditionally, understanding indoor space utilization in a typical design setting has been based on observation methodologies, where researchers document team interactions, space utilization and design activities using qualitative observation techniques. The authors of this paper propose a data mining driven methodology aimed at modeling the utilization of indoor design spaces using trajectory pattern data. Using indoor Radiofrequency identification (RFID) technology, researchers are able to collect trajectory data which can then be used to quantify the distribution of space usage patterns over time and predict future regions of interest. The proposed methodology consists of two phases: i) trajectory partitioning and ii) line segment clustering. For the first phase, trajectories are partitioned into line segments, based on unique user characteristics. In the second phase, a data mining clustering algorithm is employed to group line segments into different clusters based on a distance function. Since individual trajectories may exhibit similar movement patterns, the proposed methodology can help designers better understand how design spaces are utilized and how team dynamics evolve over time, depending on the specific design task being executed. A 3,500 square foot design space was used for the semester long study that included design teams supervised by teaching assistants. The results provide insight into the underutilization of certain regions of the design space and proposes directions towards an optimal design space methodology.