Scalable Mining of Common Routes in Mobile Communication Network Traffic Data

Abstract

A probabilistic method for inferring common routes from mobile communication network traffic data is presented. Besides providing mobility information, valuable in a multitude of application areas, the method has the dual purpose of enabling efficient coarse-graining as well as anonymisation by mapping individual sequences onto common routes. The approach is to represent spatial trajectories by Cell ID sequences that are grouped into routes using locality-sensitive hashing and graph clustering. The method is demonstrated to be scalable, and to accurately group sequences using an evaluation set of GPS tagged data.