Community elicitation from co-occurrence of activities

With reference only to observations of people's activities within a system, we investigated whether it is possible to determine the existence of their social relationships. As students interacted, either individually or in groups, we aimed to discover their social communities given the temporal and spatial co-occurrence of their activities and their implicit user–system interactions. To elicit those hidden communities, we developed two innovative approaches: history-based analysis, which exploits the similarity of users’ histories of engaging in certain activities, and session-based analysis, which uses a graph-based representation of concurrent users’ activity sessions. We tested and validated both approaches using a real-world dataset representing the activity logs of students using a virtual learning environment platform. The major results of our work confirm that the co-occurrence of people's activities is an emerging epiphenomenon of hidden, implicit exchanges of information in side-channel communications.