Recommendation system with cluster-based filtering of recommendations

• US 8,095,521 B2
• Filed: 03/30/2007
• Issued: 01/10/2012
• Est. Priority Date: 03/30/2007
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Chakrabarti, Kushal, Ionkov, George M., Chan, James D.
• Primary Examiner(s): Ali, Mohammad
• Assistant Examiner(s): HASAN, SYED HAROON
• Application Number:
  US11/694,720
• Publication Number:
  US 20080243637A1
• Time in Patent Office: 1,747 Days
• Field of Search: 707/737, 707/751, 707/999.003, 707/999.005
• US Class Current: 707/705
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0601 Electronic shopping [e-shopping]
• CPC Class Codes: G06Q 30/0631 Item recommendations

Abstract

Computer-implemented processes are disclosed for clustering items and improving the utility of item recommendations. One process involves applying a clustering algorithm to a user'"'"'s collection of items. Information about the resulting clusters is then used to select items to use as recommendation sources. Another process involves displaying the clusters of items to the user via a collection management interface that enables the user to attach cluster-level metadata, such as by rating or tagging entire clusters of items. The resulting metadata may be used to improve the recommendations generated by a recommendation engine. Another process involves forming clusters of items in which a user has indicated a lack of interest, and using these clusters to filter the output of a recommendation engine. Yet another process involves applying a clustering algorithm to the output of a recommendation engine to arrange the recommended items into cluster-based categories for presentation to the user.

Claims

1. A computer-implemented method, comprising:

maintaining a collection of items in association with a user, said collection comprising items rated by the user;

applying a clustering algorithm to said collection of items to subdivide the collection into multiple clusters, wherein the clusters are generated based, at least in part, on calculated distances between the items;

classifying at least a first cluster of said multiple clusters, based at least partly on the user'"'"'s ratings of items therein, as representing an area in which the user lacks interest;

classifying at least a second cluster of said multiple clusters, based at least partly on the user'"'"'s ratings of items therein, as representing an area of interest to the user;

obtaining a set of recommended items for the user from a recommendation engine;

filtering the set of recommended items based, at least in part, on distances between the recommended items and one or more of said clusters of items, wherein filtering the set of recommended items comprises filtering out at least one item from the set of recommended items based on a determination that the recommended item corresponds to an area in which the user lacks interest, said determination based at least partly on both (1) a calculated distance between the recommended item and said first cluster, and (2) a calculated distance between the recommended item and the second cluster, said filtering producing a filtered set of recommended items, said filtering performed automatically by a computer system; and

outputting the filtered set of recommended items for presentation to the user.

2. The method of claim 1, wherein the items are represented in a hierarchical browse structure, and wherein applying the clustering algorithm comprises using locations of the items in the hierarchical browse structure to calculate said distances.

3. The method of claim 2, wherein applying the clustering algorithm comprises calculating an entropy level of a cluster, and using said entropy level to determine whether to divide the cluster.

4. The method of claim 1, wherein classifying the first cluster as representing an area in which the user lacks interest comprises determining that the first cluster consists primarily of items rated by the user at less than a selected threshold rating.

5. The method of claim 1, wherein the method comprises determining whether the recommended item falls within a threshold distance of the first cluster.

6. The method of claim 1, wherein the distances between the recommended items and one or more of said clusters are calculated based on centers of said clusters.

7. The method of claim 1, wherein the collection includes items in which the user has explicitly indicated a lack of interest.

8. The method of claim 1, further comprising using at least two of the clusters to organize the filtered set of recommended items into cluster-based categories for presentation to the user.

9. A computer system programmed to perform the method of claim 1.

10. A computer-readable medium having stored thereon executable computer code that embodies the method of claim 1.

11. A computer system comprising one or more computing devices, said computer system programmed to implement:

a computer data repository that stores a collection of items associated with a user, said collection including items rated by the user;

a clustering component that applies a clustering algorithm to the collection of items to divide the collection into multiple clusters of items, said clustering component configured to use item ratings specified by the user to classify at least a first cluster of said multiple clusters as corresponding to an area in which the user lacks interest, and to classify at least a second cluster as corresponding to an area of interest to the user;

a recommendation engine that generates a set of personalized item recommendations for the user; and

a filtering component that filters out selected items from the set of personalized item recommendations based on determinations that the selected items are associated with the area in which the user lacks interest, as determined based at least partly on both (1) calculated distances between the selected items and the first cluster, and (2) calculated distances between the selected items and the second cluster.

12. The computer system of claim 11, wherein the clustering component uses browse node ancestries of the items as at least one criterion for forming said clusters of items.

13. The computer system of claim 11, further comprising a presentation component that outputs the filtered set of item recommendations for presentation to the user.

14. The computer system of claim 13, wherein the presentation component uses clusters generated by the clustering compo nent to organize the filtered set of personalized item recommendation into multiple categories for presentation to the user.

15. The computer system of claim 11, wherein the collection includes items in which the user has explicitly indicated a lack of interest.

16. The computer system of claim 11, further comprising a user interface that enables the user to assign a tag to multiple items at-a-time by tagging a selected one of said clusters, and enables the user to request item recommendations that are specific to said tag.

17. The computer system of claim 11, further comprising a user interface that enables the user to request item recommendations that are specific to a selected one of said clusters.

18. The computer system of claim 11, wherein the recommendation engine generates the set of personalized item recommendations based, at least in part, on items included in said collection.

19. The computer system of claim 11, wherein the clustering component is configured to assign a classification to each of the multiple clusters based on the user'"'"'s ratings of items in the respective cluster, and the filtering component is configured to use the classifications, in combination with calculations of distances between particular items and clusters, to determine which items to filter out of the set of personalized item recommendations.

20. Non-transitory computer storage that stores executable code modules that direct a computing system to perform a process that comprises:

generating a first item cluster that, based on item ratings specified by a user, corresponds to an area in which the user lacks interest;

generating a second item cluster that, based on item ratings specified by the user, corresponds to an area of interest to the user;

determining a set of recommended items for the user; and

filtering out at least one recommended item from the set of recommended items based at least partly on a determination that the recommended item corresponds to an area in which the user lacks interest, said determination based at least partly on calculated distances between the recommended item and the first and second item clusters, to thereby generate a filtered set of recommended items for the user.

21. The non-transitory computer storage of claim 20, wherein the process comprises calculating said distances based at least partly on respective locations of the items in a hierarchical browse structure.

22. The non-transitory computer storage of claim 20, in combination with a computer system programmed with said executable code modules to perform said process.