Analyzing Large Scale Wi-Fi Data Using Supervised and Unsupervised Learning Techniques

Abstract

With the increasing number of Wi-Fi enabled portable devices, and the ubiquitous Wi-Fi networks, analyzing multiple aspects of a population is becoming more insightful, inexpensive and non-intrusive. Network packets propagated from Wi-Fi enabled devices encapsulate spatial, spatiotemporal and behavioral information about the device holders. An opportunity that was available only to online stores a decade ago. In this thesis, we propose two methods to expand the possibilities of Wi-Fi Analytics. First, we present a remote localization technique as an essential preprocessing step to enable Wi-Fi Analytics in the retail and hospitality sector by analyzing non-intrusively collected Wi-Fi packets using supervised learning. Our method is capable of estimating positions without any prior knowledge about the store plan or the antennas' location with only one off-the-shelf access point. Unlike other positioning techniques, instead of estimating a relative position of a device from an antenna, we provide an absolute position for a device as inside or outside of a venue without making any assumption about the site nor the positioned devices. Second, we present a non-intrusive technique to learn about past spatial behaviors of a population by analyzing their SSID data. The main outcome of this component is to expand our knowledge about previously visited locations of a population by collecting few network packets of the Wi-Fi enabled devices and mining the data using unsupervised learning techniques.