Mining Behavioral Groups in Large Wireless LANs

Abstract

One vision of future wireless networks is that they will be deeply integrated and embedded in our lives and will involve the use of personalized mobile devices. User behavior in such networks is bound to affect the network performance. It is imperative to study and characterize the fundamental structure of wireless user behavior in order to model, manage, leverage and design efficient mobile networks. It is also important to make such study as realistic as possible, based on extensive measurements collected from existing deployed wireless networks. In this study, using our systematic TRACE approach, we analyze wireless users' behavioral patterns by extensively mining wireless network logs from two major university campuses. We represent the data using location preference vectors, and utilize unsupervised learning (clustering) to classify trends in user behavior using novel similarity metrics. Matrix decomposition techniques are used to identify (and differentiate between) major patterns. While our findings validate intuitive repetitive behavioral trends and user grouping, it is surprising to find the qualitative commonalities of user behaviors from the two universities. We discover multi-modal user behavior for more than 60% of the users, and there are hundreds of distinct groups with unique behavioral patterns in both campuses. The sizes of the major groups follow a power-law distribution. Our methods and findings provide an essential step towards network management and behavior-aware network protocols and applications, to name a few.