Ernst Versteeg, Swisscom AG
Looking into the devil's kitchen of mobility data from mobile networks: Location calculation, data streams, privacy limitations, roaming requirements, costs and upgrade enforcements for network elements

In the analysis of anonymous data from mobile networks several factors are mission-critical: Which data can be used, the limitations of each data stream, location calculation, coverage area, privacy and costs. The first part of the presentation gives a brief overview which data is available in mobile networks, which boundary conditions we found for them and how Swisscom is trying to get to the best location calculation possible. Discussed is also the huge impact from privacy rights. I especially show how we are improving the location calculation for GSM and UMTS using antenna modeling, combining it with a 3D model of Switzerland; now also taking into account the distance from antennas, using calibrated TA for GSM and RTT for UMTS, which we calibrate soon. It results in small location areas everywhere, up till into mountain area with poor coverage, while still preserving a hit rate of 95%. I show actual results and discuss also our experience with mobile network vendors to get anonymous data out of mobile networks and selling it to interested parties.

Dr. Michael May, Fraunhofer IAIS
Reality Monitoring - be ahead with real time data

Detecting critical events or events related to damage require a swift reaction. Novel methods for analyzing mobile data and Twitter messages create a headstart of knowledge, as events cannot only be detected in real time but their development can also be monitored thereafter. With "Reality Monitoring", the Fraunhofer IAIS presents intelligent methods to support the process of decision-making and visual interpretation of data. Making mobile data accessible not only for analysis methods but also for visualization purposes, offers novel and innovative insights for location planning or traffic forecasting.

Ernst Versteeg, Swisscom AG
A way out of the devil's kitchen of mobility data: Central processing of mobile location data in a data pyramid

As shown in the first presentation the limitations from the data of mobile networks, location calculation, technical boundary conditions, privacy and high costs results in an complex mix, which is extreme difficult to master. A possible solution is presented. The architecture preserves all privacy rights and is based on a simple central multi layer data aggregation engine, which combines all available location data on multiple abstraction levels. This allows cost effective analysis for multiple clients, with tracks for several days or weeks, which is much longer than a couple of hours, the current limitation of all solutions we know about for analysis of anonymous mobile data.

Dr. Mirco Nanni, KDD Lab, ISTI-CNR
Mobility data mining meets network analysis

In this talk we summarize a few experiences in mobility data analysis that put at work methods coming from two different communities: mobility data mining, focused directly on the trajectory data that describe the mobility; and network analysis, focused (in our context) on the connections between entities (individuals or places) that such trajectories induce. Finally, we will discuss the concerns about privacy that naturally emerge in this context, together with ideas on a general approach to tackle them.

Dr. Gennady Andrienko, Fraunhofer IAIS
Visual analytics of mobile phone use data

Mobile phone use data are collected now in various forms, including call detail records, handover data, and activity records for cell towers. These data have huge potential for uncovering people's movement and activities, event detection, situation awareness, and prediction of future trends and developments. We propose a suite of visual analytics techniques for analyzing mobile phone use data. The suite includes intelligent methods for spatial, temporal, and spatio-temporal aggregation in database followed by interactive visualization, methods for detecting events and studying their spatial and temporal distribution, reconstructing flows and assessing their temporal dynamics, and interactive visual techniques that support modeling. Currently we are working on developing methods for semantic interpretation of routine movements of individuals based on calling activities.