System and method for utilizing social networks for collaborative filtering
OwnerChuck P. Lam, Murad Goekselinventors
Life of this patent
Estimating the term…
Abstract
A novel system and method of predicting a user's rating of a new item in a collaborative filtering system is described. The invention incorporates social network information in addition to user ratings to make recommendations. The distance between users in the social network is used to enhance the estimate of user similarities for collaborative filtering. The social network can be constructed explicitly by users or deduced implicitly from observed interaction between users.
Claims6
A computer-implemented method of predicting an active user's rating for an item in a recommendation system implemented on computing equipment, wherein the active user and a plurality of non-active users are members of a social network implemented on computing equipment, comprising:
- providing a similarity measure, w″a,u for each pair of users in said recommendation system;
- determining ratings for items rated by non-active users in said recommendation system;
- calculating, with the similarity measure w″a,u, by the processor, a weighted average of ratings for said item to produce the predicted active user's rating;
- wherein the similarity measure, w″a,u comprises a relationship distance in said social network between pairs of users,
- wherein an initial version of the similarity measure wa,u for active user a and non-active user u in the system comprises the relationship
- w
- a
- ,
- u
- =
- ∑
- i
-
-
- (
- r
- a
- ,
- i
- -
- r
- _
- a
- )
-
- (
- r
- u
- ,
- i
- -
- r
- _
- u
- )
- ∑
- i
-
-
- (
- r
- a
- ,
- i
- -
- r
- _
- a
- )
- 2
-
- ∑
- i
-
-
- (
- r
- u
- ,
- i
- -
- r
- _
- u
- )
- 2
- where ru,i is the rating of non-active user u who has rated item i, ra,i is the rating of active user, a, who has rated item i,
- wherein calculating a second version of the similarity measure, w′a,u for active user a and neighbor user u in the system comprises enhancing the initial version correlation coefficient w″a,u in accordance with the relationship:
- w′a,u=w′a,u, if there are more than N co-rated items
- w
- a
- ,
- u
- ′
- =
- w
- a
- ,
- u
-
-
- I
- a
- ⋂
- I
- u
-
- N
- ,
- otherwise
- where Iu is the set of items that neighbor user u has rated, Ia is the set of items that active user a has rated,
- wherein the similarity measure w″a,u is enhanced from the second version w′a,u of the similarity measure in accordance with the relationship
- w″a,uηda,u+(1−ηda,u)w′a,u
- where η is a tunable parameter and da,u is a measure of relationship distance between active user a and non-active user u in the social network.
A collaborative filtering system for predicting a user's rating for an item, wherein the user and a plurality of other users are members of a social network, comprising:
- a memory storing:
- a similarity measure w″a,u for each pair of users in said recommendation system;
- ratings for items rated by other users in said recommendation system;
- a processor for calculating a weighted average of all the ratings for the item,
- wherein the weighted average is the sum of the product of a rating and its respective correlation coefficient similarity measure divided by the sum of the correlation coefficients similarity measures to provide a predicted user rating;
- wherein the similarity measure comprises a measure of similarity in ratings between pairs of users;
- wherein the similarity measure comprises a relationship distance in the said social network between pairs of users;
- wherein the processor calculates an initial version of the correlation coefficient wa,u for user a and other user u in the system in accordance with the relationship
- w
- a
- ,
- u
- =
- ∑
- i
-
- r
- a
- ,
- i
-
- r
- u
- ,
- i
- ∑
- j
- ∈
- I
- a
-
- r
- a
- ,
- j
- 2
-
- ∑
- j
- ∈
- I
- u
-
- r
- u
- ,
- j
- 2
- where ru,i is the rating of other user u who has rated item I, ra,i is the rating of user a who has rated item I, where Ia and Iu are the set of items user a and user u have rated, respectively,
- wherein the processor enhances the initial version of coefficient wa,u by calculating the second version of the correlation coefficient, w′a,u for user a and other user u in the system in accordance with the relationship:
- w′a,u=wa,u, if there are more than N co-rated items
- w
- a
- ,
- u
- ′
- =
- w
- a
- ,
- u
-
-
- I
- a
- ⋂
- I
- u
-
- N
- ,
- otherwise
- where Iu is the set of items that other user u has rated, Ia is the set of items that active user a has rated,
- wherein the processor calculates the similarity measure w″a,u by enhancing a second version w′a,u of the similarity measure in accordance with the relationship
- w″a,u=ηda,u+(1−ηda,u)w″a,u
- where η is a tunable parameter and da,u is a measure of relationship distance between user a and other user u in the social network.
Dependencies are inferred from the claim text.