Determining whether to include items in a data feed

• US 8,370,208 B1
• Filed: 02/15/2010
• Issued: 02/05/2013
• Est. Priority Date: 02/15/2010
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Gupta, Mohit, Xu, Bin, Sun, Jiaming (Jason), Parimi, Nagender
• Primary Examiner(s): HAQ, NAEEM U
• Application Number:
  US12/705,692
• Time in Patent Office: 1,086 Days
• Field of Search: 705/26.1, 705/26.2, 705/26.25, 705/26.3, 705/26.35, 705/26.4, 705 2641- 2644, 705/26.5, 705 2661- 2664, 705/26.7, 705/26.8, 705/26.81, 705/26.82, 705/26.9, 705/27.1, 705/27.2
• US Class Current: 705/26.1
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

This disclosure relates to determining whether to include items in a feed to a referral network site. A catalog is maintained of a plurality of items sold through an electronic commerce network site, wherein at least a portion of a plurality of sales of the items originate by virtue of a referral from a referral network site. A quantity of commercial data associated with the items is stored in a data store, and a predictive model is trained with the commercial data associated with the items. Each of the items is applied to the predictive model to determine whether such items are to be included in a feed to the referral network site. The feed that includes a subset of the items is sent to the referral network site.

Claims

1. A non-transitory computer readable medium embodying a program executable by at least one computing device, comprising:

an item feed application configured to periodically send a feed of a subset of a plurality of items to a comparison shopping network site over a network, the items being sold to a plurality of customers through an electronic commerce network site over a network, wherein at least a portion of a plurality of sales of the items through the electronic commerce network site originate by virtue of a referral from the comparison shopping network site;

a portion of the item feed application including;

code that stores a quantity of commercial data associated with the items in a data store, the commercial data including sales history data associated with the items;

code that trains a predictive model in the computing device with the commercial data associated with the items;

code that applies each of the items to the predictive model to determine whether such items are to be included in a feed to the referral network site; and

code that sends the feed that includes a subset of the items to the referral network site.

2. The non-transitory computer readable medium embodying the program executable by the at least one computing device of claim 1, wherein the predictive model comprises a random forest model.

3. A method, comprising the steps of:

maintaining a catalog of a plurality of items sold through an electronic commerce network site implemented in a computing device, wherein at least a portion of a plurality of sales of the items originate by virtue of a referral from a referral network site;

storing a quantity of commercial data associated with the items in a data store;

training a predictive model in the computing device with the commercial data associated with the items;

applying each of the items to the predictive model to determine whether such items are to be included in a feed to the referral network site; and

sending the feed that includes a subset of the items to the referral network site.

4. The method of claim 3, wherein the steps of training and applying are performed automatically.

5. The method of claim 3, wherein the step of training the predictive model in the computing device with the commercial data associated with the items further comprises the step of training the predictive model with a quantity of predefined events occurring in a plurality of different past time periods.

6. The method of claim 3, wherein the predictive model comprises a random forest model.

7. The method of claim 5, further comprising the step of tracking a number of selections of an item detail page for each of the items by customers during interaction with the electronic commerce network site, wherein the events further comprise the selections.

8. The method of claim 7, wherein at least some of the selections occur by virtue of the referral from the referral network site.

9. The method of claim 5, further comprising the step of tracking the sales of each of the items through the electronic commerce network site, wherein the events further comprise the sales of the items.

10. The method of claim 3, further comprising the step of maintaining a taxonomy in the computing device, the taxonomy including a plurality of item categories, wherein the commercial data further includes an association between respective ones of the item categories and respective ones of the items.

11. The method of claim 3, wherein the items are associated with a product segment.

12. The method of claim 11, wherein the items associated with the product segment include features in common with each other.

13. The method of claim 3, wherein the referral network site operates on a cost-per-click basis with respect to the electronic commerce network site.

14. The method of claim 3, wherein the referral network site operates on a cost-per-action basis with respect to the electronic commerce network site.

15. A system, comprising:

at least one processor based device having a processor and a memory;

a catalog of a plurality of items stored in the memory, the items being sold through an electronic commerce network site, wherein at least a portion of a plurality of sales of the items originate by virtue of a referral from a referral network site;

commercial data associated with each the items stored in the memory; and

a feed content control process stored in the memory and executable by the processor, the feed content control process comprising;

logic that trains a predictive model with the commercial data associated with a subset of the items;

logic that applies each of the items in the subset to the predictive model to determine whether such items are to be included in a feed to the referral network site; and

logic that generates and sends the feed to the referral network site.

16. The system of claim 15, wherein the predictive model comprises a random forest model.

17. The system of claim 15, wherein the predictive model comprises a classification model.

18. The system of claim 15, wherein the subset of items are associated with a product segment.

19. The system of claim 18, wherein the items associated with the product segment include at least one similar feature.

20. The system of claim 15, wherein the logic that generates and sends the feed to the referral network site is implemented periodically.

21. The system of claim 20, wherein the logic that trains the predictive model with the commercial data associated with the subset of the items and the logic that applies each of the items in the subset to the predictive model to determine whether such items are to be included in the feed to the referral network site are both implemented before the generation and sending of a feed.

22. The system of claim 15, wherein the logic that trains the predictive model with the commercial data associated with the subset of the items further comprises logic that trains a plurality of predictive models, where each predictive model is trained using the commercial data associated with a corresponding one of a plurality of subsets of the items.

23. The system of claim 22, wherein the items associated with each of the subsets are associated with a product segment defined among all of the items in the catalog.

24. The system of claim 15, further comprising a taxonomy maintained in the memory, the taxonomy including a plurality of item categories, wherein the commercial data further includes a plurality of associations between respective ones of the item categories and respective ones of the items.

25. The system of claim 15, wherein the commercial data includes a plurality of predefined events associated with respective ones of the items occurring within a plurality of predefined past time periods.

26. The system of claim 25, wherein the predefined events further comprise a plurality of selections of item detail pages made by customers during interaction with the electronic commerce network site, wherein each item detail page includes information about a respective one of the items.

27. The system of claim 26, wherein at least some of the selections occur by virtue of the referral from the referral network site.