On Friday, March 9, 2012 9:28:06 PM UTC, Sam Ritchie wrote:This is exactly the sort of question we could crack open by having
standard taps into large public datasets. I think that the logic is easy at
this point; I personally find myself writing queries for data I don't yet
have, or for very small, contrived datasets, rather than asking questions
of the data itself.
What (publicly available) datasets can we think of that would help us
start approaching this problem? On the twitter side, I'll start poking
around to see if it's possible to get some of our data out there.
--
Sam Ritchie
Sent with Sparrow <
http://www.sparrowmailapp.com/?sig>
On Friday, March 9, 2012 at 11:26 AM, Ted Dunning wrote:
On Fri, Mar 9, 2012 at 5:55 AM, Paul Lam wrote:
I've just thought of this although it's probably not quick enough to do in
a session. I'd love to investigate the threshold of collective action for
major riots (e.g. Tunisia) using Twitter/social network feeds.
1. When is the event-horizon of information flow in a riot. At what point
is the flow of information snowballed until action is practically
guaranteed based on network effects. This could be modelled by
considering
information flow quantities at change-point of third, fourth, or fifth
derivatives against time.
Great problem. Bad approach.
Events that are inherently counts are not good candidates for derivatives.
It is quite reasonable to do change-point detection using a hierarchical
Poisson model, however. The idea is that you have a Poisson process with a
rate that is a sample from a random walk of some kind. The details of the
walk aren't real important in that there are lots of good alternatives to
pick from. Some alternatives can be a step that occurs at random intervals
with weakly constrained long-tailed distribution or a t-distributed random
walk that takes steps very often.
The underlying rate of the Poisson process is the hidden variable that you
are looking for. Large steps in that rate are the events that you seek.
2. What is the average influence threshold until action on a personal level
I can't comment on this except that you need to model the external
information flow as well. It isn't just buddies.
Once we can model one, we might be able to model for many. Identify
patterns across the spectrum, and maybe lead to predictive capabilities for
predicting social unrests. Which has never been done before, massively
--
You received this message because you are subscribed to the Google Groups
"cascading-user" group.
To post to this group, send email to cascading-user@googlegroups.com.
To unsubscribe from this group, send email to
cascading-user+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/cascading-user?hl=en.On Friday, March 9, 2012 9:28:06 PM UTC, Sam Ritchie wrote:This is exactly the sort of question we could crack open by having
standard taps into large public datasets. I think that the logic is easy at
this point; I personally find myself writing queries for data I don't yet
have, or for very small, contrived datasets, rather than asking questions
of the data itself.
What (publicly available) datasets can we think of that would help us
start approaching this problem? On the twitter side, I'll start poking
around to see if it's possible to get some of our data out there.
--
Sam Ritchie
Sent with Sparrow <
http://www.sparrowmailapp.com/?sig>
On Friday, March 9, 2012 at 11:26 AM, Ted Dunning wrote:
On Fri, Mar 9, 2012 at 5:55 AM, Paul Lam wrote:
I've just thought of this although it's probably not quick enough to do in
a session. I'd love to investigate the threshold of collective action for
major riots (e.g. Tunisia) using Twitter/social network feeds.
1. When is the event-horizon of information flow in a riot. At what point
is the flow of information snowballed until action is practically
guaranteed based on network effects. This could be modelled by
considering
information flow quantities at change-point of third, fourth, or fifth
derivatives against time.
Great problem. Bad approach.
Events that are inherently counts are not good candidates for derivatives.
It is quite reasonable to do change-point detection using a hierarchical
Poisson model, however. The idea is that you have a Poisson process with a
rate that is a sample from a random walk of some kind. The details of the
walk aren't real important in that there are lots of good alternatives to
pick from. Some alternatives can be a step that occurs at random intervals
with weakly constrained long-tailed distribution or a t-distributed random
walk that takes steps very often.
The underlying rate of the Poisson process is the hidden variable that you
are looking for. Large steps in that rate are the events that you seek.
2. What is the average influence threshold until action on a personal level
I can't comment on this except that you need to model the external
information flow as well. It isn't just buddies.
Once we can model one, we might be able to model for many. Identify
patterns across the spectrum, and maybe lead to predictive capabilities for
predicting social unrests. Which has never been done before, massively
--
You received this message because you are subscribed to the Google Groups
"cascading-user" group.
To post to this group, send email to cascading-user@googlegroups.com.
To unsubscribe from this group, send email to
cascading-user+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/cascading-user?hl=en. 
