Mining Groups in Mobile Monitoring Log
- 1. On Mining Mobile Users by Monitoring Logs Dmitry Namiot Lomonosov Moscow State University i-ASC 2014
- 2. Dmitry Namiot http://servletsuite.blogspot.com • Passive monitoring for mobile users lets us anonymously collect presence information about mobile visitors • This information is linked to some predefined place • For any such place we can talk about some visiting patterns • How can we restore some of the patterns from our monitoring log? What are we talking about?
- 3. Dmitry Namiot http://servletsuite.blogspot.com Agenda • Passive monitoring for mobile users • Web Log analogue • Missed records and the specifics for mobile statistics • Related works • Group visits
- 4. Dmitry Namiot http://servletsuite.blogspot.com Passive monitoring • source address (MAC- address) • SSID • supported rates • additional request information • extended support rates • vendor specific information
- 5. Dmitry Namiot http://servletsuite.blogspot.com Passive monitoring • Wi-Fi router • Detects Wi-Fi (Bluetooth) devices • External database (MySQL) • 70% detection rate
- 6. Dmitry Namiot http://servletsuite.blogspot.com Web Log • Remote IP address – MAC address • User-Agent header – parsed from MAC • Missed URI field • Missed Referrer field • New field: SSID. PNL – preferred networks list
- 7. Dmitry Namiot http://servletsuite.blogspot.com Specifics • Detection rate: 70%-80% • It could not be predicted. Depends on mobile OS, applications, etc. • A reasonable assumption: the percentage for missed records is about the same • Use relative values instead of absolute figures. E.g., trend in attendance versus visitors counting • Testing hypotheses about the results of external influences
- 10. Dmitry Namiot http://servletsuite.blogspot.com Groups • Group of friends, which meets within a certain time • Not all of them are present at each meeting • Not all of them arrive simultaneously • Can we discover such groups?
- 11. Dmitry Namiot http://servletsuite.blogspot.com Clusters Increased interval Increased frequency
- 12. Dmitry Namiot http://servletsuite.blogspot.com Groups mining • find clusters for the each day • detect the sequences of clusters across all days with some minimum set of common members
- 13. Dmitry Namiot http://servletsuite.blogspot.com Conclusion • A new model for mining mobile monitoring log • Business-oriented reports about mobile groups • Tested on real example (café in office building, 8 groups from 11) • Applied areas: Smart Cities applications, retail
- 14. Dmitry Namiot http://servletsuite.blogspot.com OIT Lab • Faculty of Computational Mathematics and Cybernetics, Lomonosov Moscow State University. Research areas are: • telecom and software services, open API for telecom, Smart Cities, M2M applications, context-aware computing.