Offloading surrogates characterization via mobile crowdsensing

Abstract

This paper uses data mining of a mobile crowdsensed dataset of passive WiFi scans to define attributes that can characterize a chaotic WiFi deployment with respect to offloading opportunities. Besides indicators of signal quality, we define indicators of contact windows and contact opportunities with an Access Point (AP). We apply k-means clustering to identify classes of APs, and observe that interference metrics are more relevant than plain RSSI; that contact window metrics can be estimated using only APs’ coverage data; and that popularity and importance can characterize APs whether the offloading targets many or only a few users.