Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Automated Land Use Identification using Call Detail Records

1,130 views

Published on

3rd ACM Int. Workshop on Hot Topics in Planet-Scale Measurement, in conjuntion with ACM MobiSys2011

  • Be the first to comment

  • Be the first to like this

Automated Land Use Identification using Call Detail Records

  1. 1. <ul><li>Automated Land Use Identification
  2. 2. using Cell-phone Records </li></ul><ul>June 28, 2011 </ul><ul>Víctor Soto & Enrique Frías-Martínez </ul><ul>TELEFÓNICA I+D </ul>
  3. 3. <ul></ul><ul>Introduction </ul><ul>01 </ul><ul>Telefónica I+D </ul>
  4. 4. <ul></ul>
  5. 5. <ul>Goal: Land use of urban areas using Call Details Records. <li>Study Evolution, Evaluate Urban Zooning </li></ul><ul></ul>
  6. 6. <ul></ul><ul>Preliminaries </ul><ul>02 </ul><ul>Telefónica I+D </ul>
  7. 7. <ul>Cell Phone Network </ul><ul></ul>
  8. 8. <ul>CDR dataset </ul><ul><li>Our Dataset </li></ul><ul><ul><li>1 month of phone call interactions.
  9. 9. 1100 Base Transceiver Stations.
  10. 10. Each CDR contains: </li><ul><li>phone Source | phone Destiny | bts Source | bts Destiny | DD/MM/YYYY | hh:mm:ss | d </li></ul><li>Phone number are encrypted to anonymize user identities. </li></ul></ul><ul></ul>
  11. 11. <ul></ul><ul>Activity Signature </ul><ul>03 </ul><ul>Telefónica I+D </ul>
  12. 12. <ul>Representations </ul><ul></ul>
  13. 13. <ul></ul><ul>Land Use Identification </ul><ul>04 </ul><ul>Telefónica I+D </ul>
  14. 14. <ul></ul>
  15. 15. <ul>Methodology (I) </ul><ul><li>K-means was applied for k={3,4,...,8} for the three representations.
  16. 16. Validity index: maximizes the minimum inter-cluster distance and minimizes the average intra-cluster distance. </li></ul><ul><li>DTW also used but did not return good results. </li></ul><ul></ul>
  17. 17. <ul></ul>
  18. 18. <ul></ul><ul>Validation </ul><ul>06 </ul><ul>Telefónica I+D </ul>
  19. 19. <ul></ul>
  20. 20. <ul>Cluster 1: Industrial & Office </ul><ul></ul>
  21. 21. <ul>Cluster 2: Business & Commercial </ul><ul></ul>
  22. 22. <ul>Cluster 3: Nightlife </ul><ul></ul>
  23. 23. <ul>Cluster 4: Leisure </ul><ul></ul>
  24. 24. <ul>Cluster 5: Residential </ul><ul></ul>
  25. 25. <ul></ul><ul>Classification </ul><ul>06 </ul><ul>Telefónica I+D </ul>
  26. 26. Classification <ul><li>Cluster representatives can be used as class labels.
  27. 27. Proposed classification scheme: the class label that minimize the euclidean distance between a BTS signature and itself is assigned as the class of the area.
  28. 28. We validate the classification against the city of Barcelona: </li><ul><li>900 BTS towers.
  29. 29. Extension 100 km 2 . </li></ul></ul>
  30. 30. Classification: BCN
  31. 31. <ul></ul><ul>Conclusions & Future Work </ul><ul>07 </ul><ul>Telefónica I+D </ul>
  32. 32. <ul><li>Specific uses for City Halls. </li></ul><ul></ul>
  33. 33. “ Robust Land Use Characterization of Urban Landscapes using Cell Phone Data” V. Soto, E. Frias-Martinez The First Workshop on Pervasive Urban Applications (PURBA), in conjunction with the Ninth International Conference on Pervasive Computing in San Francisco, CA, USA on June 12-15, 2011. www.enriquefrias-martinez.info/publications [email_address]

×