Abstract
The increase in mobile devices available and their significant part of our daily routine have resulted in location data being very accessible, up to date and usually very accurate.
Access to this data enables understanding of human's location behavior in terms of both time and place. An advanced aspect of such understanding is being able to recognize patterns in human mobility. Location Pattern Recognition identifies common routes that users preform taking into consideration time, geographical location and the semantic meaning of the location.
Recognizing location patterns can be a force multiplier to different fields and applications: traffic and road planning, parking allocation, life style services, public services, recommendation systems and more. For example: Based on users location patterns one can foresee the amount of traffic in a certain road junction, during a specific hour of the day.
Using such prediction can enable adjusting the traffic lights in the junction accordingly, and preventing possible traffic jams.
The goal of our research is to present methods for detecting Location Patterns of users and evaluating the different methods. A location patterns can take into consideration different variables: time, geographic location, semantic location, single users and multiple users. All of the above are in the scope of our work.
In our research we analyze cellular location data of thousands of user located in Tel-Aviv area during a period of two weeks. We use a battery of models that can represent location patterns such as Markov Models, Bayesian Networks and Variable -Order Bayesian Networks. Initial results, including single users' Location Patterns, have been obtained.
