Cryptography Reference
In-Depth Information
trust value, the reputation of a user can be used as insight into the user's relevance to
recommendation. Trust weighting calculates a trust value for every user by cross-
validation using equation (3).
m
correct
uvi
,,
(3)
tru st
=
i
=
1
ui
,
m
recom m end
uvi
,,
i
=
1
The system then computes the prediction. To incorporate values from the trust model
into recommendation, the system filters the trust value with some threshold value [3].
The robustness of relevance weighting is evaluated as item-trust and similarity using
equation (4).
2
*
sim
*
trust
u
,
i
v
,
i
w
=
(4)
u,v,i
sim
+
trust
u
,
i
v
,
i
where sim u,v is Pearson's correlation coefficient. A prediction for the target user is
computed using equation (2), replacing sim u,v with w u,v,i .
4 Experimental Results of the Modified RPCF Algorithm
In the experiment, the hotel ratings datasets from Travelocity is used to evaluate
RPCF algorithm. The dataset contains 2721 ratings from 40995 users' reviews for 740
hotels. Each user can rate a hotel to express his/her willingness to stay at this hotel
and a rating is a number ranging from 1 to 5. A higher score indicates a higher
preference.
Table 2. Hotel Rating Profiles showing Push Attack and Nuke Attack
A Push Attack Favoring H10
A Nuke Attack Favoring H12
Users
H1 H2 H3 … H10 … H15 H1 H2 H3 … H12 … H15
Alice
5
2
5
?
4
5
2
5
?
4
U1
5
3
3
2
5
5
3
3
4
5
U2
4
3
2
3
4
3
2
5
U3
5
4
2
4
5
4
4
4
U4
4
3
1
4
4
3
5
4
⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞ ⁞
Att 1
1
5
1
1
Att 2
2
1
5
3
2
1
1
3
Att 3
1
2
5
1
2
1
Att 4
1
1
5
2
1
1
1
2
Att 5
1
3
5
1
3
1
Search WWH ::




Custom Search