Abstract
Using a pedestrian tracking system on laser scanners, the big data for office-workers' trajectories can provide detailed deep insights on workplace designs. In the analysis of tracking data, it is necessary to represent them on maps so that the relationships between spatial and temporal features can be understood clearly to comprehend characteristics of their working style. In this paper, two methods to extract patterns of “Spatio-temporal activity” are described. One is the time zone extraction model, i.e. a classification model on the basis of information loss minimization model, and the other is the day scene extraction model, i.e. a latent class model using the probabilistic latent semantic indexing (PLSI). Numerical studies demonstrate the usefulness of our proposed models.
