what-when-how
In Depth Tutorials and Information
and found a regular reading relation (RR relation) between social relations and
readership relations. hey deined the weblog regular reading relation as follows:
Regular Reading (RR): Blogger
A
has an RR relation with blog
B
if blogger
A
reads blog
B
more frequently than every
m
times the blogger logs in.
he experiment results show that if
m
= 5, half of the blogs that are read more
than three times. Moreover, they use five attributes to predict the creating and
holding of reading links between two weblogs (
x
and
y
) based on a machine-learn-
ing approach:
◾
Adamic/Adar:
∑
∈
Γ
1
/ log |
( ) |
z
∩
z
Γ
(
x
)
Γ
(
y
)
◾
Gragh distance: Length of shortest path between
x
and
y
◾
Common neighbors:
|
Γ Γ
( )
x
∩
( ) |
y
◾
Jaccard's coefficient:
|
Γ Γ
Γ Γ
( )
x
∩
( ) | / |
y
( )
x
∪
( ) |
y
◾
Preferential attachment:
⎮Γ
(
x
)
⎮⋅⎮Γ
(
y
)
⎮
where
Γ
(
x
) is the set of neighbors of
x
in the network. hese attributes are built
based on four social networks: citation, blogroll, comment, and trackback net-
works. he detail description and testing of these attributes are given in References
45 and 46.
3.2.2.2 Predicting the Posting Behavior Using
a Dynamic Probability Model
his is different from the read behaviors in that the user writing about something
will be influenced by more factors. Generally, the users' posting behavior is influ-
enced by their interests [37], social network link [44], group behavior [47], time
factor, outside news source, etc. Some of these factors are difficult to obtain and
analyze. For instance, we cannot understand the newspapers, news site pages, or
other news source that one user would read, unless we can track the user's life.
However, investigators have analyzed and modeled posting behavior by factors that
can be obtained from some open data sources.
Zhou et al. [47] studied the factors that influence the user's posting—behav-
ior—whether one user will post something about one specific topic or not and pro-
posed a dynamic probability model that could predict the tendency of the user's
posting behavior in the online social network. In the topic discussion process,
a user's posting behavior would be mostly affected by three factors: individual
interest, group behavior, and time lapse. he relative hypotheses are proposed as
follows:
Hypothesis. 1:
Individual interest factor. Assuming that the more times one
user attends the discussion on topic
T
at present and has attended in the
past, the more probably it is that he or she will attend the discussion the
next time.
