Information Technology Reference
In-Depth Information
relatively one-sided, they should also be take the direct effect on node influence which
is given by behavior between nodes in account. The behavior of forward and comment
can change the size of the impact of a user. Therefore, when measure of influence of
nodes we should fully consider the interaction between the nodes that play an
important role for influence.
We proposed model of influence rank (model of IR).The method in this paper
proposed by combining the interactive behavior between nodes and node's attributes.
Firstly we need to give a node attribute's quantification. Node attribute has many
factors; these factors will influence effect of node. We should select some main
influence factors, and then we use analysis hierarchy process to give the weight of
every factor of influence, using the weighted to calculate the quantification of node
attribute value. Secondly we introduce the thought of PageRank [13] algorithm to
study the node interaction behavior. The PageRank algorithm is based on the
assumption: the webpage is more important when it link to more webpage; webpage is
more important when it linked to the more important webpage. Similarly, for online
social networks, when the user is more commented by others, its influence will be
greater; when the user is forward comments by more important users; its influence will
be greater. Finally the user attributes and user behavior as the node influence factor
synthesis node's influence in online social networks.
4
The Evaluation of Node's Influence
4.1
Measuring Node Attributes
There are lots of node's attributes in online social network; we select some attributes
which have more obvious role: user type, numbers of fans, numbers of forwarded,
numbers of attention, numbers of micro-blog, and numbers of comments. The numbers
of fans, numbers of forwarding, and the numbers of comments can reflect the influence
of nodes from different aspects. The analytic hierarchy process to solve the weight
problem of each influence factor, and then use the weighted and calculate the
quantization node attribute value. Calculation steps as follows:
Step1:
Construction of index matrix
X
and normalization, so we get new matrix
A
as follows:
xx
x
aa
a
11
12
1
m
11
12
1
n
xx
x
aa
a
21
22
2
m
X
=
21
22
2
n
A
=
xx
x
aa
a
n
1
n
2
nm
m
1
m
2
mn
We can use the standard 0-1 transform to normalize:
min
xx
−
(1)
ij
j
a
=
ij
max
x
−
x
j
ij
Where
max
j
the maximum value in the j column is,
min
j
is the minimum value in the
j
x
x
column.
