BEHAVIORAL FILTER FOR PERSONALIZED RECOMMENDATIONS

• US 20130013448A1
• Filed: 09/15/2012
• Published: 01/10/2013
• Est. Priority Date: 01/27/2011
• Status: Active Grant

• Current Assignee: Andrew J. Bradley, David Neil Turner, Matthew J. Round, Oliver Ruper Masting, Robert Philip Japp
• Original Assignee: Andrew J. Bradley, David Neil Turner, Matthew J. Round, Oliver Ruper Masting, Robert Philip Japp
• Inventors: Round, Matthew J., Bradley, Andrew J., Turner, David Neil, Japp, Robert Philip, Masting, Oliver Ruper
• Granted Patent: US 8,473,369 B2
• Time in Patent Office: Days
• US Class Current: 705/26.7
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions

Abstract

This disclosure describes systems and associated methods that can selectively filter seed behavior, e.g., user activity used to generate item recommendations. In some embodiments, seed behaviors and catalog items are associated with categories in an electronic catalog, and a particular seed behavior is used to generate user recommendations if it is more recent than a user'"'"'s last purchase in the category of the seed behavior. For example, a user'"'"'s activity in the TV category, e.g., viewing various TV models, may not be used to generate recommendations if the activity occurred prior to the user'"'"'s purchase of a TV. As a result, in certain embodiments, additional TVs may not appear in the user'"'"'s recommendations following her purchase, reflecting that the user has fulfilled her desire to purchase a TV.

Claims

1-20. -20. (canceled)

21. One or more physical computer storage devices comprising a recommendations system configured to generate personalized recommendations, the recommendations system comprising computer-executable instructions that cause a computing system to at least:

access user activity data of a target user to identify an item purchased by the target user, the purchased item being represented by item data in an electronic catalog, the user activity data comprising data representing item selection activity of the user with respect to the electronic catalog;

identify purchase-related activity of the target user from the user activity data, the purchase-related activity being related to the purchase and activity occurring prior to a purchase of the purchased item by the target user;

identify non-purchase-related activity of the target user from the user activity data, the non-purchase-related activity occurring subsequent to the purchase and therefore being unrelated to the purchase;

apply first weights to the non-purchase-related activity;

apply second weights to the purchase-related activity, the second weights being lower than the first weights;

generate recommendations for the target user based at least in part on the user activity data of the target user, wherein said generation of the recommendations comprises ranking the recommendations based on the first weights applied to the non-purchase-related activity and the second weights applied to the purchase-related activity, such that at least some of the recommendations derived from the non-purchase-related activity are ranked higher than at least some of the recommendations derived from the purchase-related activity; and

selecting a subset of the recommendations based on the rankings to present to the target user.

22. The one or more physical computer storage devices of claim 21, wherein said identification of the purchase-related activity comprises identifying items selected from the electronic catalog by the target user that are classified in a category of the purchased item.

23. The one or more physical computer storage devices of claim 22, wherein said generation of the recommendations further comprises using one or more media items selected by the target user to generate the recommendations despite the one or more media items being from the same category as the category of the purchased item.

24. A method of generating personalized recommendations, the method comprising:

accessing user activity data of a target user to identify an item purchased by the target user, the purchased item being represented by item data in an electronic catalog, the user activity data comprising data representing item selection activity of the user with respect to the electronic catalog;

identifying purchase-related activity of the target user from the user activity data, the purchase-related activity being related to the purchase and activity occurring prior to a purchase of the purchased item by the target user;

identifying non-purchase-related activity of the target user from the user activity data, the non-purchase-related activity occurring subsequent to the purchase and therefore being unrelated to the purchase;

applying weights to the non-purchase-related activity;

applying second weights to the purchase-related activity, the second weights being lower than the first weights; and

generating recommendations for the target user based at least in part on the user activity data of the target user and the first and second weights, such that at least some of the recommendations derived from the purchase-related activity are weighted lower than at least some of the recommendations derived from the purchase-related activity.

25. The method of claim 24, wherein said identifying the purchase-related activity comprises identifying items selected from the electronic catalog by the target user that are classified in a category of the purchased item.

26. The method of claim 25, wherein said generating the recommendations further comprises using one or more media items selected by the target user to generate the recommendations despite the one or more media items being from the same category as the category of the purchased item.

27. A system for generating personalized recommendations, the system comprising:

a seed behavior collector configured to access user activity data of a target user to identify a purchased item purchased by the target user, the purchased item comprising an item represented in an electronic catalog;

a related history filter implemented in computer hardware, the related history filter configured to identify, from the user activity data, purchase-related activity data of the target user, the purchase-related activity data being related to the purchase and occurring prior to the purchase, and identify second activity data of the user activity data occurring subsequent to the purchase; and

a recommendation engine configured to provide recommendations for the target user based at least in part on the user activity data of the target user, the recommendation engine further configured to filter out at least some of the recommendations generated based on the purchase-related activity data of the user activity data.

28. The system of claim 27, wherein the related history filter is further configured to identify the purchase-related activity by at least identifying items selected from the electronic catalog by the target user that are classified in a category of the purchased item.

29. The system of claim 28, wherein the recommendation engine is further configured to generate the recommendations by at least using one or more media items selected by the target user to generate the recommendations despite the one or more media items being from the same category as the category of the purchased item.

30. The system of claim 27, wherein the system is implemented by a plurality of computing devices.