System and method for utilizing social networks for collaborative filtering

• US 20050256756A1
• Filed: 05/17/2004
• Published: 11/17/2005
• Est. Priority Date: 05/17/2004
• Status: Active Grant

• Current Assignee: Chuck P. Lam, Murad Goeksel
• Original Assignee: Chuck P. Lam, Murad Goeksel
• Inventors: Lam, Chuck P., Goeksel, Murad
• Granted Patent: US 7,689,452 B2
• Time in Patent Office: Days
• US Class Current: 705/10
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: H04N 21/25891 being end-user preferences retrieval of video data in a video database based on user preferences G06F16/739; arrangements for recognizing users' preferences H04H60/46; user profiles in network data switching protocols H04L67/306; processing of user preferences or user profiles in wireless networks H04W8/18
• CPC Class Codes: H04N 21/4662 characterized by learning algorithms
• CPC Class Codes: H04N 21/4788 communicating with other users, e.g. chatting arrangements for providing for computer conferences, e.g. chat rooms, to substation in data switching networks H04L12/1813; distributed application using peer-to-peer [P2P] networks H04L67/104
• CPC Class Codes: H04N 21/4826 using recommendation lists, e.g. of programs or channels sorted out according to their score

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.

Claims

1. A method of predicting a user'"'"'s rating for an item in a recommendation system, comprising:

providing a similarity measure for each pair of users in said recommendation system;

determining ratings for items rated by other users in said recommendation system;

calculating a weighted average of ratings for said item;

wherein the plurality of users are members of a social network;

wherein the similarity measure comprises a relationship distance in said social network between pairs of users.

2. The method of claim 1, wherein the social network is a network graph of relationships between people, wherein the social network is constructed out of explicit or implicit social relationships, wherein the degrees of separation between people within the social network is a primary metric of distance between the nodes of the network.

3. The method of claim 1, wherein the predicted user rating of an item i for an active user a is calculated in accordance with the relationship:

4. The method of claim 3, wherein the initial version of the correlation coefficient (also known as the Pearson correlation coefficient;

w_a,u) for active user a and neighbor user u in the system comprises the relationship

5. The method of claim 3, wherein the second version of the correlation coefficient, w′

_a,u for active user a and neighbor user u in the system, further comprising;

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

$w_{a,u}^{\mathrm{\prime <br><br>}}=w_{a,u}<br><br>\frac{<br><br>I_a\bigcap <br><br>I_u<br><br>}{N},$

otherwise

where I_u is the set of items that neighbor user u has rated, I_a is the set of items that active user a has rated, N is the number of co-rated items between the active user a and the neighbor user u.

6. The method of claim 3, wherein the third version of the correlation coefficient, w″

_a,u for active user a and neighbor user u in the system, further comprising;

enhancing the second version of the correlation coefficient w′

_a,u in accordance with the relationship;

w″

_a,u=η

^da,u+(1−

η

^da,u)w′

_a,u

where tunable parameter η

controls the level of emphasis given to close friends, d_a,u is a measure of distance between active user a and user neighbor user u in the social network or alternatively d_a,u can be other measures of distance on a social network.

7. A collaborative filtering system for predicting a user'"'"'s rating for an item, comprising:

a memory storing;

a similarity measure for each pair of users in said recommendation system;

ratings for items rated by other users in said recommendation system;

a processor for calculating the 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 divided by the sum of the correlation coefficients to provide a predicted user rating;

wherein the plurality of users are members of a social network;

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.

8. The system of claim 7, wherein the social network is a network graph of relationships between people, wherein the social network is constructed out of explicit or implicit social relationships, wherein the degrees of separation between people within the social network is a primary metric of distance between the nodes of the network.

9. The system of claim 7, wherein the processor calculates the predicted user rating of an item i for a active user a in accordance with the relationship:

10. The system of claim 7, wherein the processor calculates the initial version of the correlation coefficient (also known as the Pearson correlation coefficient;

w_a,u) for active user a and neighbor user u in the system which comprises the relationship

11. The system of claim 7, wherein the processor enhances the initial version of coefficient w_a,u by calculating the second version of the correlation coefficient, w′

_a,u for active user a and neighbor user u in the system in accordance with the relationship;

w′

_a,u=w_a,u, if there are more than N co-rated items

$w_{a,u}^{\mathrm{\prime <br><br>}}=w_{a,u}<br><br>\frac{<br><br>I_a\bigcap <br><br>I_u<br><br>}{N},$

otherwise

where I_u is the set of items that neighbor user u has rated, I_a is the set of items that active user a has rated, N is the number of co-rated items between the active user a and the neighbor user u.

12. The system of claim 7, wherein the processor enhances the second version of coefficient w′

_a,u by calculating the third version of the correlation coefficient, w″

_a,u for active user a and neighbor user u in the system, in accordance with the relationship;

w″

_a,u=η

^fa,u)w′

_a,u

where tunable parameter T7 controls the level of emphasis given to close friends, d_a,u is a measure of distance between active user a and user neighbor user u in the social network or alternatively d_a,u can be other measures of distance on a social network.