Mining Changes in Mobility Patterns From Smartphone Data

Determining the behaviour and habits of individuals and the structure and dynamics of organizations needs human activity modeling from large-scale sensor data. The data can be derived from sensors embedded in smartphones, which have become an integral part of our daily lives. The analysis of the smartphone data provides valuable information about the users' daily routines and also helps us to find any unexpected changes. This research proposes a method to detect the changes and anomalies in a user's mobility pattern based on smartphone data. The data includes connectivity from cell tower ID (CID), wifi, Bluetooth, and any other embedded sensors in smartphone. We model the data from each day as a point in a feature space and then define a metric to find the similarities between the points. Therefore, each point represents user's mobility in a day and close points denote similar behaviours. The main challenge is defining the appropriate features which abstract mobility patterns in the corresponding day. One of the features is temporal clusters which show when the user changes his location.