Human mobility from GSM data - A valid alternative to GPS?

Characteristics of human mobility are a valuable source of information in many applications. In this paper we evaluate the usability of call detail records for the extraction of mobility quantities. We derive several quantities from the simultaneously collected GPS and GSM mobility data of the Nokia Mobile Data Challenge. Our analyses show that GSM activity data underestimates average daily travel distance and radius of gyration when derived straightforward from the data. In addition, they indicate that the correlation between mobile phone usage and movement quantities is biased when using GSM activity data. Finally, our analyses confirm that long-term GSM activity data is well suited to detect frequent stop locations.