BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Predicting Mobile User Location through Nonlinear Time Series Anal
 ysis - Salvatore Scellato (University of Cambridge)
DTSTART:20090616T120000Z
DTEND:20090616T124500Z
UID:TALK18737@talks.cam.ac.uk
CONTACT:Stephen Kell
DESCRIPTION:Accurate and fine-grained prediction of user location  and the
 ir geographical\nprofile has interesting applications including targeted c
 ontent and\nadvertisement dissemination\, recreational social network tool
 s and many more in \nareas such as psychology and anthropology.  Existing 
 techniques based on linear and probabilistic models are not able to provid
 e accurate prediction of the\nmovement patterns from a spatio-temporal per
 spective\, since they cannot capture \nthe nonlinear characteristics of th
 e behavior of the users if present. \n\nA contribution of this paper is th
 e identification of some degree of\ndeterminism\, previously uncaptured\, 
  in patterns of visits of humans to specific\nplaces\, at least for the sc
 enarios taken into consideration in this work\, based\non user GPS positio
 n datasets and base station registration data.  We then \nillustrate an ap
 proach to location prediction based on nonlinear time series\nanalysis of 
 the arrival and residence times of users in relevant places that are\nauto
 matically extracted by mining their movement patterns.  Moreover\, we repo
 rt\nabout our  evaluation over these datasets which confirms a prediction 
 accuracy\nwhich ranges between 65% and 90% even after a number of hours.  
 We compare our\nforecasting results to those obtained by means of the pred
 iction techniques\nproposed in the literature\, showing we have more stabl
 e accuracy over time. We\nalso report the performance of an application fo
 r dissemination of contents that\nare characterized by spatio-temporal con
 straints. 
LOCATION:Computer Laboratory\, William Gates Building\, Room FW11
END:VEVENT
END:VCALENDAR