Method and system for cold-start item recommendation

Patent metadata

Field	Value
Publication number	US-10699198-B2
Application number	US-201414519273-A
Country	US
Kind code	B2
Filing date	Oct 21, 2014
Priority date	Oct 21, 2014
Publication date	Jun 30, 2020
Grant date	Jun 30, 2020

Abstract

Method, system, and programs for estimating interests of a plurality of users with respect to a new piece of information are disclosed. In one example, historical interests of the plurality of users are obtained with respect to one or more existing pieces of information. One or more users are selected from the plurality of users. Historical interests of the one or more users can minimize an objective function over the plurality of users. Interests of the one or more users are obtained with respect to the new piece of information. Estimated interests of the plurality of users are generated with respect to the new piece of information based on the obtained interests of the one or more users.

First claim

We claim:

1. A method, implemented on a machine having at least one processor, storage, and a communication platform connected to a network for estimating interests, the method comprising:
generating a historical interest vector for each of a plurality of users with respect to one or more existing pieces of information;
identifying a similarity/dissimilarity between the plurality of users based on the generated historical interest vectors;
selecting a first group of users from the plurality of users, the selecting including:
determining a plurality of user sets, wherein each user set comprises a same predetermined number of users,
generating a plurality of function values by calculating an objective function based on vectors for each of the plurality of user sets,
selecting a user set from the plurality of user sets so that the function value generated based on the user set is least among the plurality of function values, and
iteratively eliminating a user from the user set based on the generated historical interest vectors of other users in the user set minimizing the objective function over a second group of users;

detecting interactions of each user in the selected first group of users with respect to a new piece of information to obtain interest of the user in the new piece of information; and
generating estimated interests of the second group of users with respect to the new piece of information based on the detected interactions.

2. The method of claim 1 , wherein a number of users included in the first group of users is predetermined.

3. The method of claim 1 , wherein the objective function represents an expected mean square error between the estimated interests and real interests of the second group of users with respect to the new piece of information.

4. The method of claim 1 , wherein selecting the first group of users from the plurality of users comprises:
initializing the user set to comprise the plurality of users;
generating a matrix which initially comprises a plurality of columns, wherein each column of the matrix corresponds to a generated vector for one of the users in the user set;
generating a plurality of candidate matrices each of which corresponding to a user in the user set and generated by removing the column corresponding to the user from the matrix;
generating a plurality of function values by calculating the objective function based on each of the plurality of candidate matrices;
selecting one of the plurality of candidate matrices so that the function value generated based on the selected candidate matrix is least among the plurality of function values;
updating the matrix with the selected candidate matrix; and
updating the user set by removing the user corresponding to the selected candidate matrix from the user set.

5. The method of claim 1 , wherein the estimated interests of the plurality of users are generated based on a least squares model.

6. The method of claim 1 , further comprising:
identifying a user other than the plurality of users; and
estimating interest of the user based on the obtained interests of the first group of users.

7. The method of claim 1 , wherein the first group of users is distinct with respect to the second group of users.

8. The method of claim 1 , wherein the second group of users is different than the first group of users.

9. A system, including a storage, and a communication platform connected to a network for estimating interests, the system comprising:
at least one processor configured to
generate a historical interest vector for each of the plurality of users with respect to one or more existing pieces of information;
identify a similarity/dissimilarity between the plurality of users based on the generated historical interest vectors
select a first group of users from the plurality of users, the selecting including:
determining a plurality of user sets, wherein each user set comprises a same predetermined number of users,
generating a plurality of function values by calculating an objective function based on vectors for each of the plurality of user sets,
selecting a user set from the plurality of user sets so that the function value generated based on the user set is least among the plurality of function values, and
iteratively eliminating a user from the user set based on the generated historical interest vectors of other users in the user set minimizing the objective function over a second group of users;
detect interactions of each user in the selected first group of users with respect to a new piece of information to obtain interest of the user in the new piece of information; and
generate estimated interests of the second group of users with respect to the new piece of information based on the detected interactions.

10. The system of claim 9 , wherein a number of users included in the first group of users is predetermined.

11. The system of claim 9 , wherein the objective function represents an expected mean square error between the estimated interests and real interests of the second group of users with respect to the new piece of information.

12. The system of claim 9 , wherein the at least one processor is further configured to:
initialize the user set to comprise the plurality of users;
generate a matrix which initially comprises a plurality of columns, wherein each column of the matrix corresponds to a generated vector for one of the users in the user set;
generate a plurality of candidate matrices each of which corresponding to a user in the user set and generated by removing the column corresponding to the user from the matrix;
generate a plurality of function values by calculating the objective function based on each of the plurality of candidate matrices;
select one of the plurality of candidate matrices so that the function value generated based on the selected candidate matrix is least among the plurality of function values;
update the matrix with the selected candidate matrix; and
update the user set by removing the user corresponding to the selected candidate matrix from the user set.

13. The system of claim 9 , wherein the estimated interests of the plurality of users are generated based on a least squares model.

14. The system of claim 9 , wherein the at least one processor is further configured to:
identify a user other than the plurality of users; and
estimate interest of the user based on the obtained interests of the first group of users.

15. A machine-readable tangible and non-transitory medium having information recorded thereon for estimating interests, wherein the information, when read by the machine, causes the machine to perform the following:
generating a historical interest vector for each of a plurality of users with respect to one or more existing pieces of information;
identifying a similarity/dissimilarity between the plurality of users based on the generated historical interest vectors;
selecting a first group of users from the plurality of users, the selecting including:
determining a plurality of user sets, wherein each user set comprises a same predetermined number of users,
generating a plurality of function values by calculating an objective function based on vectors for each of the plurality of user sets,
selecting a user set from the plurality of user sets so that the function value generated based on the user set is least among the plurality of function values, and
iteratively eliminating a user from the user set based on the generated historical interest vectors of other users in the user set minimizing the objective function over a second

Assignees

• Yahoo Holdings Inc
• Oath Inc

Inventors

• Somekh Oren Shlomo
• Golan Shahar
• Golbandi Nadav
• Karnin Zohar
• Rokhlenko Oleg
• Anava Oren
• Lempel Ronny

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title

• G06Q30/0631

  Recommending goods or services · CPC title

• G06N7/005

  Physics · mapped topic