Detecting expectation-based spatio-temporal clusters formed during opportunistic sensing

Detecting clusters in the encounter graphs generated from reality mining data is one way of detecting the social and spatial relationships of participants. However, many of the existing clustering algorithms do not factor in the time since encounters, and can only be used to describe a single aggregated snapshot of the data. This paper describes a spatio-temporal clustering technique which has been used to reveal the transient communities within the data.