Information Technology Reference
In-Depth Information
λ
−
m
CI
=
max
(5)
m
−
1
CI
<
0.1
If
,we can accept the weight of different indexes. Otherwise, we need to
recalculate the
Step2
.
w
and the index
Step3:
Calculating AR according to the weight of each index
i
a
:
value of each node
m
i
=
1, 2,
(6)
,
n
AR i
()
=
a w
ij
j
j
=
1
4.2
Behavior Measurement
In this process, we use PageRank algorithm to solve the problem of interaction
between nodes.PageRank is one of the core technology of Google, its basic idea is to
determine the importance of using webpage hyperlink structure webpage, The
PageRank generates represent each page importance value becomes the PR value, a
page's PR value not only depends on the number of connected to the page, but also be
influenced by the quality and importance of the page, but the page's PR value will be
evenly distributed to its chain of the page[14].The calculating formula of PageRank as
follows:
PR p
()
(7)
j
PR p
()
=+−
d
1 )
d
i
Lp
()
pMp
∈
()
j
j
i
Where
LP
is number of
P
links to other page, where
M P
is numbers of chain
()
j
()
i
into
P
,
d
is damping coefficient.
Then extract the relationship of forwarded or comments in online social network, a
directed graph of forward comments relationship could be constructed. In the
following simple diagram, the direction of the arrow represents the object forwarding
comments. Because of the large scale network, the analysis is difficult; we can start
from a node to consider. Observation of the following figure, for example the user a
forward or comment of user e
、
f and g. Of course, there are also have other users will
forward information of user a. The user e and g may also go forward or comment on
other user, this part of the forward or comment leave out in Fig.1.We consider only the
relationship between local forward or comments, and then further promotion.
e
a
f
b
g
Fig. 1.
Schematic diagram of forwarding
