what-when-how
In Depth Tutorials and Information
G2
G3
a3
a5
Enter Group()
JOIN
a2
a4
a1
G1
Figure 5.2
A probabilistic social group evolution model; after JOIN action.
(From H.-C. Chen et al., Personalization inferring agent dynamics from social
communicationnetworks,in
International Conference on Knowledge Discovery
and Data Mining
,2007,pp.36-45.)
higher the probability that the user will attend the same discussion in the future.
Based on the assumption that each user has his or her own unique interest in a par-
ticular discussion topic contents, if one user has attended the same discussion often
in over a period of time in the past, then that user would be much more interested
in that particular topic. herefore, the user would be much more likely to discuss
the topic again in the future and attend that same discussion in the future.
he second hypothesis states that, through the group behavior factor, it can be
assumed that if the group of users that attend a discussion about topic T increases,
then there will be a higher probability that one of the users will attend the discus-
sion again in the future. his hypothesis is similar to the irst in many respects
and, in some ways, is simply an addition to it with different characteristics. It also
can be stated similarly that if the group of users develop their own unique interests
and that group grows larger, then the group will be much more likely to revisit the
discussion and continue it in the future.
he third and final hypothesis states that, through the time lapse factor, it can
be assumed that the longer the interval between the present and peak time becomes,
the lower the probability that users will attend the discussion in the future. his
hypothesis continues with the concepts of interest in topics and expands to include
the whole-time factor. Sensibly, the more the time between topic discussions
attended, the less memorable the topic will become to the user and, therefore, the
less interest will be shown in it. As interest lowers for a topic, the probability of
the user attending a discussion on this same topic grows smaller, as stated in the
hypothesis.
Based on the first hypothesis, a behavior tendency function can be developed,
f
(
-
). his function takes input variables of
a
and time
n
. he function calculates
