Inferring the activities of smartphone users from context measurements using Bayesian inference and random utility models

Abstract Smartphones collect a wealth of information about their users' environment and activities. This includes GPS (global positioning system) tracks and the MAC (media access control) addresses of devices around the user, and it can go as far as taking visual and acoustic samples of the user's environment. We present a Bayesian framework for the identification of the current activity type of a smartphone user. As the prior information, we use a random utility model that predicts the type of activity a user is likely to perform given (i) the user's socioeconomic features, (ii) the land use of the user's current location, and (iii) the time of day. This model is estimated using data from the 2005 Swiss transport microcensus. The smartphone measurements come from a experimental survey, where one user carried around a phone programmed to constantly record his GPS location and other context variables, including the MAC addresses of nearby Bluetooth devices. In addition to this, the user answered a daily survey, where he described and geo-located all the activities performed during this period. An analysis of the recorded data shows that the information about nearby Bluetooth devices can be related to particular activities that are conducted jointly with the owners of that devices. The likelihood function is therefore specified as the probability of observing particular Bluetooth devices when conducting particular activities. Due to the limited amount of available data, only exemplary results are given, which, however, clearly indicate that the accuracy of the prior model can be greatly improved by using Bluetooth data.