what-when-how
In Depth Tutorials and Information
=
u
min(
1
CENTRALITY CLIQUE
,
)
(11.30)
net k i
,
CLIQUE and CENTRALITY are the normalized clustering coefficient and central-
ity, respectively.
k
k
=
O
'
O
net
O
(11.31)
i
i
j
j
k
i
O k i ' and O j k i are the modified opinion and original opinion of node k i about node
j, respectively.
k
k
k n
4 O O
=
⊕ ⊕
(11.32)
i
'
O
'
...
O
'
1
2
j
j
j
Algorithm of updating trust relationship [25]: Assume that node i and node j have
existed in the network for a long time so that they can update their trust relationship
based on certain experiences. he algorithm below illustrates the updating process.
Find all links from node i to node j in the network. L
= {
}
L L
,
,...,
L n
is a set
1
2
= {
}
of all links where L
=
i p p
,
,
,...,
p
,
j
p
p p
,
,...,
p
is a set of all nodes
i
1
2
m
1
2
m
in the link.
k
= {
}
1 , ,..., is a set of all nodes that have opinion about node j and are not
included in the set P . We can then measure the network context in the algorithm
of establishing trust relationship. And we can determine:
k k
k l
L
p
p
p
=
O
'
O
i
O
O
...
O
i
1
2
m
j
j
p
p
p
11.33)
1
2
3
where O L i
represents the opinion passing along the link.
'
k
k
k
L
L
L n
(11.34)
O O
=
O
⊕ ⊕
...
O
O
O
⊕ ⊕
...
O
i
'
'
'
'
'
'
1
2
l
1
2
j
j
j
j
j
j
j
he model presented in [25] also has some shortcomings. he subjective logics are
not appropriate enough to model uncertain probabilities. Besides, the model assumes
that the recommendation sources are equally reliable. hus, more complex operators
are required in the trust model, which will lead to increasing time complexity.
11.4.9 Trusted Gossip Protocol
In Reference 27 is proposed a trusted gossip protocol that is used to maximize the spread
of acceptable stories while simultaneously minimizing the spread of unacceptable sto-
ries by a story filtering process. It is very useful in the rating feedback system. Message
level filtering can be used as supplementary for node level recommendation.
Search WWH ::




Custom Search