Chatty Mobiles:Individual mobility and communication patterns

Human mobility analysis is an important issue in social sciences, and mobility data are among the most sought-after sources of information in ur-ban studies, geography, transportation and ter-ritory management. In network sciences mobility studies have become popular in the past few years, especially using mobile phone location data [1,2,3,4,5]. For preserving the customer pri-vacy, datasets furnished by telecom operators are anonymized. At the same time, the large size of datasets often makes the task of calculating all observed trajectories very difficult and time-con-suming. One solution could be to sample users. However, the fact of not having information about the mobile user makes the sampling deli-cate. Some researchers [1] select randomly a sample of users from their dataset. Others try to optimize this method, for example, taking into account only users with a certain number or fre-quency of locations recorded [2,3]. At the first glance, the second choice seems to be more effi-cient: having more individual traces makes the analysis more precise. However, the most fre-quently used CDR data (Call Detail Records) have location generated only at the moment of communication (call, SMS, data connection). Due to this fact, users’ mobility patterns cannot be precisely built upon their communication pat-terns. Hence, these data have evident short-comings both in terms of spatial and temporal scale. In this paper we propose to estimate the correla-tion between the user’s communication and mo-bility in order to better assess the bias of fre-quency based sampling. Using technical GSM network data (including communication but also independent mobility records - as in [6]), we will analyze the relationship between communication and mobility patterns.