Pervasive Infrastructure Ce ll Phone N e t w ork Cell Phone networks are built using Base Transceiver Stations (BTS). Each BTS will be characterized by a feature vector that describes the calling behavior area. 1
Pervasive Infrastructure CDR da t a se t Our Dataset • 1 month of phone call interactions. • 1100 Base Transceiver Stations. • Each CDR contains: › phoneSource | phoneDestiny | btsSource | btsDestiny | DD/MM/YYYY | hh:mm:ss | d • Phone number are encrypted to anonymize user identities. Traffic M b o ility alg rith s o m Subscribers sample 2233445566|15/02/ 2008| 2233445567|15/01/ 2008| 2233445568|15/07/ 2008|25/07/2010 2233445569|15/09/ 2008| Cell catalogue 1
Hotspot Detection• What is a hotspot? – In this context a hotspot is understood as a concentration of people (or activities) over a specific period of time and a specific geographic area.• Interesting for urban planning, emergency relief, public health, context‐aware services• Approach – Greedy clustering algorithm seeded with local maxima – Hotspots based on activity or on number of people.
Hotspot Detection• Data: – CDR from Mexico for a period of 4 months. • Output: – At a national level: cities. At an urban level: city blocks. Evolution of dense areas for urban planning.
Land Use Classification• Aggregate and clean data for each BTS. – Obtain signature of each BTS (total number of calls every hour: 24 hours average week day and 24 hours average weekend day) – BTS based Voronoi gives the tessellation for land classification. – Automatic Identification of clusters with similar behaviour that maximize the compactness of the groups identified.
Land Use ClassificationR e p r e s e n t a t io n s Activity signature vectors are built: each component contains the number of managed calls by the BTS in 5-minute intervals. 1
Land Use Classification• Industrial Parks / Office Areas
Land Use Classification• Commercial ‐ Residential
Conclusiones• Traditional approaches are costly and based on questionnaires.• Urban Dynamics can be modelled using pervasive infrastructures• Reduction in cost, increment of the flexibility• Possibility of real‐time modelling