Presentation 10th Feb 2010

February 11th, 2010 Leave a comment Go to comments

Gave presentation to Paddy, Davide, Neil Cowzer and Fergal Reid (clique) about my quick and dirty analysis of the dataset that I have collected allready.

Slides

General concensus was that there was not really enough users, and so there were some suggestions about other datasets that might be found -persuade a mobile phone company to give data about user movements. Mine flickr/twitter for geo-tagged photo’s/tweets, and try to determine groups of people based on similar locations.

Also suggested that the GMA is good for visualising data, not greatly interesting, PH is interesting as is SPD. BD is something that is useful as an application to gather data, but would need a very large engineering effort.

Paddy suggested that if we could make the data collection process very easy, then we could throw it out to the student population to start collecting data. Fergal said that in J2ME it would be very difficult, but by sticking to C++ it might work (for Nokia phones).

Also talked about getting ground truth for data, Fergal Suggested collecting accellorometer data too (so if someone asked – how did you verify GPS trace, one can say that we correlated it with the accelorometer data). I also suggested tagging locations.

Determined the following actions:

  • Look for access to datasets with good location – 1 week
    • WaveLAN Dataset
    • HeaNET – chase paddy – Eduroam
    • Mine location data from Flickr
  • Look at applying analysis to these datasets – specifically
    • Periodicity Hunting
    • Spatial Dependance on the Degree
  • See if we can construct overlay over these networks
    • e.g. drop nodes
      • Popular locations
      • popular people
      • Other?
      • Vector clocks might be the way to do it
  • Read up about Vector Clocks as suggested in the paper by Klineberg, Watts and ???? at  KDDOA
  • Speak to Graham about whether I can easily integrate this data into his code, if so – do it, otherwise think about implementing it seperately(robustly!)

Also planned to meet Paddy again next week to go over these things, and try to hammer out a better plan. Then meet with these people again in three weeks to show where I have go to.

Davide also talked about churn in proximity patterns – might be worth thinking about what this means (example was then a person regularly sees other people, and after a while, one of those people drops off the radar – what does this mean)

Paddy said that in his mind, the long goal is to be able to forward plan using the knowledge of data that has passed (prediction).

  1. No comments yet.
  1. No trackbacks yet.