ADAPTIVE RANKING OF NEWS FEED IN SOCIAL NETWORKING SYSTEMS

• US 20130031034A1
• Filed: 07/29/2011
• Published: 01/31/2013
• Est. Priority Date: 07/29/2011
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Gubin, Max, Kao, Wayne, Vickrey, David, Maykov, Alexey
• Granted Patent: US 8,768,863 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 10/06393 Score-carding, benchmarking or key performance indicator [KPI] analysis
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

Machine learning models are used for ranking news feed stories presented to users of a social networking system. The social networking system divides its users into different sets, for example, based on demographic characteristics of the users and generates one model for each set of users. The models are periodically retrained. The news feed ranking model may rank news feeds for a user based on information describing other users connected to the user in the social networking system. Information describing other users connected to the user includes interactions of the other users with objects associated with news feed stories. These interactions include commenting on a news feed story, liking a news feed story, or retrieving information, for example, images, videos associated with a news feed story.

Claims

1. A method comprising:

determining a plurality of sets of users of a social networking system, each set characterized by values of attributes describing the users;

for each set of users;

training a model configured to rank news feed stories for presentation to users of the set, the training utilizing training sets obtained from the set of users, wherein features used for the model comprise interactions of users of the set with news feed stories presented to the users;

ranking stories identified for presentation to a user using the model and presenting the ranked stories to the user; and

periodically retraining the model.

2. The method of claim 1, wherein the model is periodically retrained at a rate determined using the interactions of users of the set with the newsfeed stories

3. The method of claim 2, wherein the rate of periodically retraining the model is further determined using interactions of each user with objects associated with newsfeed stories, the objects represented in the social networking system.

4. The method of claim 1, wherein periodically retraining a model for a set comprises:

receiving a first rate of interactions with news feed stories from users of the set, the first rate measured over a first time interval;

receiving a second rate of interactions with news feed stories from users of the set, the second rate measured over a second time interval occurring after the first time interval; and

responsive to the second rate indicating more than a threshold change compared to the first rate, retraining the machine learning model.

5. The method of claim 1, wherein the features used for the model further comprise interactions of users of the set with objects associated with news feed stories presented to the users

6. The method of claim 5, wherein an object associated with a news feed story comprises one of an image, a video, a comment from a user, or a user profile.

7. The method of claim 1, wherein the rate of periodically training the model is determined based on a change in the rate of interactions with newsfeed stories by the users of the set.

8. The method of claim 1, wherein a set of users is characterized by attributes describing demographic information of the users.

9. The method of claim 1, wherein a set of users is characterized by information describing connections of the users.

10. The method of claim 1, wherein a set of users is characterized by number of connections of the users.

11. The method of claim 1, wherein a set of users is characterized by a rate at which each user interacts with the social networking system.

12. The method of claim 1, wherein a set of users is characterized by a length of time interval each user of the set has interacted with the social networking system.

13. The method of claim 1, wherein a first model corresponding to a first set of users is trained using a first set of features and a second model based on a second set of users is trained using a second set of features.

14. The method of claim 1, wherein an interaction with a news feed story comprises requesting additional information associated with the news feed story.

15. The method of claim 1, wherein an interaction actions with a news feed story comprises one of selecting a link in the news feed story, commenting on the news feed story, liking the news feed story, or hiding the news feed story.

16. A method for generating a machine learning model for ranking news feed stories of a social networking system, the method comprising:

training a model configured to rank news feed stories for presentation to a set of users of a social networking system;

for each user of the set;

identifying news feed stories for presentation to the user;

using the model to rank the identified news feed stories; and

presenting the ranked news feed stories to the user;

observing interactions of users with news feed stories presented to the users; and

periodically retraining the model.

17. The method of claim 16, wherein the rate of periodically retraining the model is further determined using interactions of each user with objects associated with newsfeed stories, the objects represented in the social networking system.

18. The method of claim 17, wherein an object associated with a news feed story comprises one of an image, a video, a comment from a user, status messages, external links, content generated by the social networking system, applications, games, or user profile.

19. The method of claim 16, wherein the rate of periodically training the model is determined based on a change in the rate of interactions with newsfeed stories by the users.

20. The method of claim 16, wherein an interaction with a news feed story comprises requesting additional information associated with the news feed story.

21. The method of claim 16, wherein an interaction actions with a news feed story comprises one of selecting a link in the news feed story, commenting on the news feed story, liking the news feed story, or hiding the news feed story.

22. A computer program product having a computer-readable storage medium storing computer-executable code for ranking news feed stories of a social networking system, the code comprising a model for ranking:

a machine learning module configured to;

determine a plurality of sets of users of a social networking system, each set characterized by values of attributes describing the users;

for each set of users;

train a model configured to rank news feed stories for presentation to users of the set;

a news feed manager module configured to

for each user of the set;

identify news feed stories for presentation to the user;

use the model to rank the identified stories; and

present the ranked news feed stories to the user;

receive information describing actions associated with the news feed stories presented, the actions performed by users of the set; and

the machine learning module further configured to;

periodically retrain the model.