GDC: Group Discovery Using Co-location Traces

Abstract

Abstract-Smart phones can collect and share Bluetooth co-location traces to identify ad hoc or semi-permanent social groups. This information, known to group members but otherwise unavailable, can be leveraged in applications and protocols, such as recommender systems or delay-tolerant forwarding in ad hoc networks, to enhance the user experience. Group discovery using Bluetooth co-location is practical because: (i) Bluetooth is embedded in nearly every phone and has low battery consumption, (ii) the short wireless transmission range can lead to good group identification accuracy, and (iii) privacy-conscious users are more likely to share co-location data than absolute location data. This paper proposes the Group Discovery using Co-location traces (GDC) algorithm, which leverages user meeting frequency and duration to accurately detect groups. GDC is validated on one month of data collected from 141 smart phones carried by students on our campus. Users rated GDC's groups 30% better than groups discovered using the K-Clique algorithm. Additionally, GDC lends itself more easily to a distributed implementation, which achieves similar results with the centralized version while improving user's privacy.

Human interaction discovery in smartphone proximity networks

“…The overall complexity of Algorithm 1 is O(KLT ) where K is the number of sampling iterations (we set K = 100 in our experiments). Compared to previous works [38,25] for which the complexity grows superlinearly (quadratically or sometimes exponentially) with the problem size, GroupUs scales well with the number of links and the number of interaction types, and hence it can learn from large-scale data in linear time.…”

“…Eagle et al [10] analyed a network constructed from BT links and phone call logs to identify friendship networks. Mardenfeld et al [25] also studied a BT network to discover groups. Other works have relied on other mobile sensors, like infrared, and microphones, to address the limitations of Bluetooth to detect real face-to-face proximity, rather than just detecting people sharing an office or a large space [15,40,30].…”

TIE: Temporal Interaction Explorer for Co-presence Communities

“…With this in mind we designed a new algorithm, Group Discovery using Co-location Traces (GDC) [8]. It uses copresence traces to identify groups.…”

“…The use of such validation techniques is demonstrated by Mardenfeld, et al [8], where an after-study survey revealed the GDC algorithm finds measurably better groups than another popular community detection algorithm, KClique. This survey does not help evaluate alternative group detection algorithms, illustrating the need for some other way to evaluate these results.…”

Crowdsensing: Socio-Technical Challenges and Opportunities

“…Crowdsensing permits a huge number of sensing devices that share the collected data by the purpose to enumerate a phenomena of mutual interest [1]. Mobile devices are equipped with different sensors such as camera, GPS, digital compass, microphone, light sensor, accelerometer, and bluetooth as proximity sensor [2]. Crowdsensing empowers a large amount of mobile phones to be utilized for trading data among their clients, as well as for activities which might have an enormous societal impact.…”