Abstract

How much can the places we visit in our daily life reveal about us? Is it possible to successfully assert a subject's location and affiliation via the data obtained from cell phones? In the present research we present a cascade binary classifier framework to infer a subject's location and affiliation. In the first stage, the user location is obtained by combining clustering and a location classifier. In the second stage, the contextual spatio-temporal data is used to discriminate the user affiliation. This framework is tested using a dataset of actual cellphone usage and bluetooth encounters. Three binary classifiers are used in each stage to validate the proposed method. Encouraging results are obtained showing a high classification success rate across the whole user base.