Mining of user event data to identify users with common interests

• US 8,224,773 B2
• Filed: 11/07/2011
• Issued: 07/17/2012
• Est. Priority Date: 03/30/2005
• Status: Expired due to Fees

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Spiegel, Joel R.
• Primary Examiner(s): Cottingham, John
• Assistant Examiner(s): Arjomandi, Noosha
• Application Number:
  US13/290,859
• Publication Number:
  US 20120047150A1
• Time in Patent Office: 253 Days
• Field of Search: 707/999.2, 707/999.3, 707/706, 707/609, 707/802, 709/224
• US Class Current: 707/609
• CPC Class Codes: G06F 16/24 Querying
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/08 Auctions
• CPC Class Codes: G06Q 50/01 Social networking
• CPC Class Codes: H04L 12/1813 for computer conferences, e.g. chat rooms instant messaging H04L51/04; protocols for multimedia communication H04L65/1101; arrangements for multi-party communication H04L65/403; telephonic conference arrangements H04M3/56; television conference systems H04N7/15
• CPC Class Codes: H04L 51/52 for supporting social networking services
• CPC Class Codes: H04L 67/306 User profiles

Abstract

A computer-implemented matching service matches users to other users, and/or to user communities, based at least in part on a computer analysis of event data reflective of user behaviors. The event data may, for example, evidence user affinities for particular items represented in an electronic catalog, such as book titles, music titles, movie titles, and/or other types of items that tend to reflect the traits of users. Event data reflective of other types of user actions, such as item-detail-page viewing events, browse node visits, search query submissions, and/or web browsing patterns may additionally or alternatively be considered. By taking such event data into consideration, the matching service reduces the burden on users to explicitly supply personal profile information, and reduces poor results caused by exaggerations and other inaccuracies in such profile information.

Claims

1. A computer-implemented method of matching users to other users, the method comprising:

for each of a plurality of users of an electronic catalog, storing, in computer storage, event data representing user-generated events that reflect user affinities for particular items reflected in the electronic catalog;

programmatically generating a score that reflects a degree to which item affinities of a first user of said plurality of users are similar to item affinities of a second user of said plurality of users, said score taking into consideration a first plurality of items corresponding to the item affinities of the first user and a second plurality of items corresponding to the item affinities of the second user, wherein generating the score comprises weighting a first item and a second item identified in both the first and second plurality of items, wherein the first and second items are different, wherein the first and second items are weighted differently based at least in part on a first inherent characteristic of the first item and a second inherent characteristic of the second item, and wherein the first and second inherent characteristics are different; and

based at least in part on the score, programmatically determining whether to recommend the second user to the first user.

2. The method of claim 1, wherein generating the score further comprises according greater weight to items that are deemed to be reflective of user traits.

3. The method of claim 1, wherein generating the score further comprises according greater weight to items that are ordered by both users.

4. The method of claim 1, wherein the score is reflective of a degree to which respective search histories of the first and second users are similar.

5. The method of claim 1, wherein the event data identifies at least one of the following:

(a) items purchased by particular users, (b) items rented by particular users, (c) items added to an electronic shopping cart by particular users, (d) items added to a rental queue by particular users, (e) items rated by particular users, (f) items selected for viewing by particular users, (g) search queries submitted by particular users to search the electronic catalog.

6. The method of claim 1, wherein the event data identifies at least two of the following:

(a) items purchased by particular users, (b) items rented by particular users, (c) items added to an electronic shopping cart by particular users, (d) items added to a rental queue by particular users, (e) items rated by particular users, (f) items selected for viewing by particular users, (g) search queries submitted by particular users to search the electronic catalog.

7. The method of claim 6, wherein generating the score comprises taking into consideration at least two types of event data reflective of user affinities for items represented in the electronic catalog.

8. The method of claim 1, wherein the type of event data corresponding to the first item is different from the type of event data corresponding to the second item.

9. The method of claim 1, wherein the method is performed in response to a search request submitted by the first user.

10. The method of claim 1, further comprising, in response to a determination that the second user is to be recommended to the first user, electronically providing to the first user information sufficient for the first user to contact the second user.

11. The method of claim 10, further comprising electronically providing to the first user information about items ordered by the second user.

12. The method of claim 1, wherein weighting a first item and a second item identified in both the first and second plurality of items, wherein the first and second items are different, wherein the first and second items are weighted differently based at least in part on a first inherent characteristic of the first item and a second inherent characteristic of the second item, and wherein the first and second inherent characteristics are different comprises:

identifying items having a first inherent characteristic ordered in common between the first and second plurality of items;

identifying items having a second, different inherent characteristic ordered in common between the first and second plurality of items; and

weighting each of the identified common items as a function of the respective inherent characteristic of the identified common item.

13. The method of claim 1, wherein the weights associated with the respective inherent characteristics of the first and second items are determined based on a prioritization provided by the first user with respect to the first and second inherent characteristics.

14. The method of claim 1, wherein the weights associated with the respective inherent characteristics of the first and second items are specified by the first user.

15. A computer-implemented method of matching users to other users, the method comprising:

for each of a plurality of users, collecting and storing, in computer storage, event data representing user-generated events reflective of user behaviors, said event data collected via a computer network without requiring the users to explicitly provide personal preference information;

programmatically generating a score that reflects a degree to which behaviors of a first user of said plurality of users are similar to behaviors of a second user of said plurality of users, said score taking into consideration a first plurality of items corresponding to the behaviors of the first user and a second plurality of items corresponding to the behaviors of the second user, wherein generating the score comprises identifying items ordered in common between the first and second plurality of items and weighting each of the identified items as a function of an inherent characteristic of each item, wherein the identified items comprise a first item having a first inherent characteristic and a second item having a second inherent characteristic, and wherein the first and second items are weighted differently based at least in part on the first inherent characteristic of the first item and the second inherent characteristic of the second item; and

based at least in part on the score, programmatically determining whether to recommend the second user to the first user.

16. The method of claim 15, wherein the score reflects a degree to which the first plurality of items and the second plurality of items are related.

17. The method of claim 15, further comprising, in response to a determination that the second user is to be recommended to the first user, electronically providing to the first user information sufficient for the first user to contact the second user.

18. The method of claim 15, wherein the weights associated with the respective inherent characteristics of the first and second items are determined based on a prioritization provided by the first user with respect to the first and second inherent characteristics.

19. The method of claim 15, wherein the weights associated with the respective inherent characteristics of the first and second items are specified by the first user.

20. A computer-implemented method of matching users to other users, the method comprising:

for each of a plurality of users, storing, in computer storage, event data representing item selection histories;

programmatically generating a score that reflects a degree to which item selections of a first user of said plurality of users are similar to item selections of a second user of said plurality of users, said score taking into consideration a first plurality of items corresponding to items selected by the first user and a second plurality of items corresponding to items selected by the second user, wherein generating the score comprises weighting a first item and a second item identified in both the first and second plurality of items, wherein the first and second items are different, wherein the first and second items are weighted differently based at least in part on a first inherent characteristic of the first item and a second inherent characteristic of the second item, and wherein the first and second inherent characteristics are different; and

based at least in part on the score, programmatically determining whether to recommend the second user to the first user.

21. The method of claim 20, wherein the event data representing item selection histories is reflective of user browsing activities across a plurality of independent web sites.

22. The method of claim 21, wherein the score is reflective of a degree to which the web site affinities of the first and second users are similar.

23. The method of claim 20, wherein the score additionally reflects personal preference infomation explicitly supplied by at lease one of the first and second users.