Recommending media items based on purchase history

• US 10,097,664 B2
• Filed: 04/26/2013
• Issued: 10/09/2018
• Est. Priority Date: 04/26/2013
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Ralston, David E., Sherman, Daniel Joseph, Bell, Christopher Laurence, Priyadarshan, Eswar, Williams, Andrew
• Primary Examiner(s): Choudhury, Azizul Q
• Application Number:
  US13/871,573
• Publication Number:
  US 20140324965A1
• Time in Patent Office: 1,992 Days
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: H04L 67/306 User profiles

Abstract

A media item recommendation can be based on application purchase data. Application purchase data of a population of users can be used to create application preference clusters that represent unique application preference profiles. Media preference data of the same population of users can be used to create media preference clusters that represent unique media preference profiles. A preference relationship between an application preference cluster and a media preference cluster can be identified based on the number of members of the media preference cluster that are also members of the application preference cluster. Upon receiving a recommendation request from a user, the user'"'"'s application preference profile can be determined based on the user'"'"'s application purchase data. A media item can then be recommended based on a media preference cluster that has a preference relationship with an application preference cluster that represents the application preference profile of the user.

Claims

1. A method comprising:

receiving, at a server computer and via a network, a recommendation request requesting an electronic media item be served to a first electronic account;

determining, by the server computer, that music preference data associated with the first electronic account is unable to recommend an electronic music item;

obtaining, by the server computer, a plurality of application preference clusters that are each associated with an electronic application preference profile, wherein the application preference clusters are created by a first clustering operation that clusters a first population of electronic accounts into the application preference clusters based on software applications purchased online by each of the first population of electronic accounts;

selecting from the application preference clusters and by the server computer, a first application preference cluster associated with a first electronic application preference profile for the first electronic account, wherein the first electronic application preference profile corresponds to the first electronic account at an application preference strength level determined from similarities between the first electronic application preference profile and applications purchased by the first electronic account;

obtaining, by the server computer, a plurality of music preference clusters that are each associated with an electronic music preference profile, wherein the music preference clusters are created by a second clustering operation that clusters a second population of electronic accounts into music preference clusters based on online music preference data for the second population of electronic accounts;

identifying, by the server computer, a preference relationship between the first application preference cluster and a first music preference cluster, wherein the preference relationship is associated with a preference relationship strength value that indicates a likelihood that an account within the first application preference cluster is also part of the first music preference cluster;

identifying, by the server computer, a first electronic music preference profile for the first music preference cluster to be used for recommending electronic media items for the first electronic account based on the application preference strength level and the preference relationship strength value; and

selecting, by the server computer, at least one electronic music item to be served to the first electronic account based on the identified first electronic music preference profile.

2. The method of claim 1, wherein the identifying of the preference relationship between the first application preference cluster and the first music preference cluster includes:

determining that the first population of electronic accounts represented in a first application preference cluster over-index in the first music preference cluster.

3. The method of claim 1, wherein selecting at least one electronic music item comprises:

identifying at least one electronic music item commonly purchased by the second population of electronic accounts represented in the first music preference cluster.

4. The method of claim 1, wherein selecting the first application preference cluster associated with the first electronic application preference profile for the first electronic account comprises:

calculating a mathematical representation of the first electronic account based on application purchase data from which the software applications purchased by the first electronic account are identifiable; and

determining that the first electronic account has the first electronic application preference profile when the first application preference cluster is associated with a smallest distance measurement from the mathematical representation of the first electronic account.

5. The method of claim 2, wherein determining that the first population of accounts represented in the first application preference cluster over-index in the first music preference cluster comprises:

determining that the preference relationship strength value is greater than a predetermined number, wherein the preference relationship strength value is based on a number of electronic accounts within the first application preference cluster and the first music preference cluster.

6. The method of claim 2, wherein determining that the first population of electronic accounts represented in the first application preference cluster over-index in the first music preference cluster comprises:

determining that that the preference relationship strength value is greater than a second preference relationship strength value for the first music preference cluster and a second application preference cluster.

7. The method of claim 1, wherein the preference relationship strength value is based on a portion of the second population of electronic accounts represented by the first music preference cluster that are also a portion of the first population of electronic accounts represented by the first application preference cluster.

8. The method of claim 1, wherein the first clustering operation includes performing a centroid based clustering algorithm.

21. The method of claim 1, wherein the second cluster operation is configured to utilize the online music preference data as input into a centroid based clustering algorithm.

22. The method of claim 21, wherein the centroid based clustering algorithm is a k-means function algorithm.

23. The method of claim 1, wherein the first clustering operation and the second clustering operation each include a hierarchical based clustering algorithm.

9. A server computer comprising:

at least one processor; and

at least one memory containing instructions that, when executed, cause the at least one processor to;

receive, via a network, a recommendation request requesting an electronic target-category item be served to a first electronic account;

determine that electronic target-category preference data associated with the first electronic account is not capable of recommending an electronic music item;

determine a plurality of source-category preference clusters that are each associated with an electronic source-category preference profile, wherein the source-category preference clusters are created by a first clustering operation that clusters a first population of electronic accounts into the source-category preference clusters based on online source-category preferences of each of the first population of electronic accounts;

select from the source-category preference clusters a first source-category preference cluster associated with a first electronic source-category preference profile for the first electronic account, wherein the first electronic source-category preference profile corresponds to the first electronic account at a source-category strength level determined from similarities between the first electronic source-category preference profile and source-categories purchased by the first electronic account;

determine a plurality of target-category preference clusters that are each associated with an electronic target-category preference profile, wherein the target-category preference clusters are created by a second clustering operation that clusters a second population of electronic accounts into target-category preference clusters based on online target-category preference data for the second population of electronic accounts;

identify a preference relationship between the first source-category preference cluster and a first target-category preference cluster, wherein the preference relationship is associated with a preference relationship strength value that indicates a likelihood that an account within the first source-category preference cluster is also part of the first target-category preference cluster;

identify a first electronic target-category preference profile for the first target-category preference cluster to be used for recommending target-category items for the first electronic account;

select at least one target-category item based on the identified first electronic target-category preference profile; and

recommend, based on the recommendation request, the selected at least one target-category item.

10. The server computer of claim 9, wherein the instructions to identify the preference relationship between the first source-category preference cluster and the first target-category preference cluster further cause the at least one processor to:

determine that the first population of electronic accounts represented in the first source-category preference cluster over-index in the first target-category preference cluster.

11. The server computer of claim 9, wherein the instructions to select at least one target-category item comprises instructions to:

identify at least one target-category item commonly purchased by the second population of electronic accounts represented in the first target-category preference cluster.

12. The server computer of claim 9, wherein the instructions to select the first source-category preference cluster associated with the first electronic source-category preference profile for the first electronic account further comprises instructions to:

calculate a mathematical representation of the first electronic account based on source-category preference data identifying source-category preferences of the first electronic account; and

determine that the first electronic account has the first electronic source-category preference profile when the first source-category preference cluster is associated with a smallest distance measurement from the mathematical representation of the first electronic account.

13. The server computer of claim 10, wherein the instructions to determine that the first population of accounts represented in the first source-category preference cluster over-index in the first target-category preference cluster comprises instructions to:

determine that the preference relationship strength value is greater than a predetermined number, wherein the preference relationship strength value is based on a number of electronic accounts within the first source-category preference cluster and the first source-category preference cluster.

14. The server computer of claim 10, wherein the instructions to determine that the population of electronic accounts represented in the first source-category preference cluster over-index in the first target-category preference cluster comprises instructions to:

determine that the preference relationship strength value is greater than a second preference relationship strength value for the first target-category preference cluster and a second source-category preference cluster.

15. At least one non-transitory computer-readable medium comprising instructions that, when executed by a computing device, cause the computing device to:

receive, via a network, a recommendation request requesting an electronic media item be served to a first electronic account;

determine that music preference data associated with the first electronic account is unable to recommend an electronic music item;

create, based on a first clustering operation, a plurality of application preference clusters that are each associated with an electronic application preference profile, wherein the first clustering operation clusters a first population of electronic accounts into the application preference clusters based on software applications purchased online by each of the first population of electronic accounts;

select, from the application preference clusters, a first application preference cluster associated with a first electronic application preference profile for the first electronic account, wherein the first electronic application preference profile corresponds to the first electronic account at an application preference strength level determined from similarities between the first electronic application preference profile and applications purchased by the first electronic account;

create, based on a second clustering operation, a plurality of music preference clusters that are each associated with an electronic music preference profile, wherein the second clustering operation clusters a second population of electronic accounts into music preference clusters based on online music preference data for the second population of electronic accounts;

identify a preference relationship between the first application preference cluster and a first music preference cluster, wherein the preference relationship is associated with a preference relationship strength value that indicates a likelihood that an electronic account within the first application preference cluster is also part of the first music preference cluster;

identify a first electronic music preference profile for the first music preference cluster to be used for recommending electronic media items for the first electronic account based on the application preference strength level and the preference relationship strength value; and

select at least one electronic music item based on the identified first electronic music preference profile.

16. The non-transitory computer-readable medium of claim 15, wherein the instructions further cause the computing device to:

determine that the first population of electronic accounts represented in a first application preference cluster over-index in the first music preference cluster.

17. The non-transitory computer-readable medium of claim 15, wherein the instructions to select the at least one electronic music item comprises instructions to:

identify at least one electronic music item commonly purchased by the second population of electronic accounts represented in the first music preference cluster.

18. The non-transitory computer-readable medium of claim 16, wherein the instructions to select the first electronic application preference further comprises instructions to:

calculate a mathematical representation of the first electronic account based on application purchase data from which the software applications purchased by the first electronic account are identifiable; and

determine that the first electronic account has the first electronic application preference profile when the first application preference cluster is associated with a smallest distance measurement from the mathematical representation of the first electronic account.

19. The non-transitory computer-readable medium of claim 16, wherein the preference relationship strength value is based on a portion of the second population of electronic accounts represented by the first music preference cluster that are also a portion of the first population of electronic accounts represented by the first application preference cluster.

20. The non-transitory computer-readable medium of claim 16, wherein the first clustering operation includes performing a centroid based clustering algorithm.