Computer-implemented collaborative filtering based method for recommending an item to a user

• US 6,049,777 A
• Filed: 03/14/1997
• Issued: 04/11/2000
• Est. Priority Date: 06/30/1995
• Status: Expired due to Term

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Sullivan, James J., Sheena, Jonathan Ari, Metral, Max E., McNulty, John Edward
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Kazimi, Hani M.
• Application Number:
  US08/818,515
• Time in Patent Office: 1,124 Days
• Field of Search: 705/10, 705/14, 705/16, 705/26-28, 702/179, 707/1, 707/3, 707/103
• US Class Current: 705/7.32
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0203 Market surveys; Market polls
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions
• CPC Class Codes: G06Q 30/0605 Supply or demand aggregation

Abstract

An object for providing isolated, hierarchical data storage can be used in a method for recommending an item to one of a plurality of users. The data object abstracts an associated physical memory element and provides an interface for storing data and retrieving data from the physical memory element. In some embodiments the data object is provided with an indicator for identifying another data object that is used if a memory request is unable to be serviced by the associated physical memory element. In other embodiments this data object can be used to efficiently and transparently store profile data associated with a system for recommending items to users.

Claims

1. In a computer system having a processor and a memory, the memory connected to the processor and storing computer executable instructions, a method for recommending an item to one of a plurality of users, wherein the item is not yet rated by said one user, the method comprising the steps, implemented through said instructions, of:

(a) storing a user profile, in the memory, for each of a plurality of users, wherein the user profile comprises a separate rating value, supplied by a particular one of the users, for each corresponding one of a plurality of items, said items including the item non-rated by the user;

(b) storing an item profile, in the memory, for each of the rated items, wherein the item profile comprises a separate rating value, for a particular one of the items, provided by each one of the plurality of the users, wherein the user profile and the item profile are distinct from each other;

(c) calculating, for each one of the plurality of users and in response to the user and item profiles, a plurality of similarity factors, between said each one user and at least one other one of the users, for each of said items including said non-rated item;

(d) selecting, in response to the plurality of similarity factors and for each one of the plurality of users, a plurality of neighboring ones of the users, such that each of the neighboring ones of the users has an associated similarity factor which is greater than a first predefined threshold value or, if a confidence factor is associated with the associated similarity factor, both the associated similarity factor is less than the first predefined threshold and the confidence factor exceeds a second predefined threshold value

(e) assigning a corresponding weight to each of the neighboring users so as to define a plurality of weights; and

(f) recommending at least one of a plurality of the items to said one user in response to the plurality of weights and ratings given to the non-rated item by the neighboring ones of the users.

2. The method of claim 1 wherein the user profile comprises either a set of user n-tuples or a first set of pointers to storage locations in the memory at which user entries are stored, wherein each of the user n-tuples comprises a separate rating value, supplied by a particular one of the users, for each corresponding one of a plurality of items, said items including the item non-rated by the user and each of the user entries stores a separate rating for an associated one of the users for a corresponding one of the items.

3. The method of claim 2 wherein the item profile comprises either a set of item n-tuples or a second set of pointers to storage locations in the memory at which item entries are stored, wherein each of the item n-tuples comprises a separate rating value, for a particular one of the items, provided by each one of the plurality of the users and each of the item entries stores a rating for an associated one of the items by a corresponding one of the users.

4. The method of claim 3 wherein the calculating step further comprises the steps of:

receiving a rating from a given one of the plurality of users for a given one of the plurality of items so as to define a received rating;

updating the user profile associated with the given one user by either writing the received rating into an appropriate one of the user n-tuples or writing the received rating into an appropriate one of the user entries;

updating the item profile associated with the given one item by writing the received rating into an appropriate one of the item n-tuples or writing the received rating into an appropriate one of the item entries; and

calculating, for the given one user, a plurality of said similarity factors, wherein each of the similarity factors for said given one user represents similarity between said given one user and another one of the users with respect to the items.

5. The method of claim 1 wherein the item profile comprises either a set of item n-tuples or a second set of pointers to storage locations in the memory at which item entries are stored, wherein each of the item n-tuples comprises a separate rating value, for a particular one of the items, provided by each one of the plurality of the users and each of the item entries stores a rating for an associated one of the items by a corresponding one of the users.

6. The method of claim 5 wherein the user profile comprises either a set of user n-tuples or a first set of pointers to storage locations in the memory at which user entries are stored, wherein each of the user n-tuples comprises a separate rating value, supplied by a particular one of the users, for each corresponding one of a plurality of items, said items including the item non-rated by the user and each of the user entries stores a separate rating for an associated one of the users for a corresponding one of the items.

7. The method of claim 6 wherein the calculating step further comprises the steps of:

receiving a rating from a given one of the plurality of users for a given one of the plurality of items so as to define a received rating;

updating the user profile associated with the given one user by either writing the received rating into an appropriate one of the user n-tuples or writing the received rating into an appropriate one of the user entries;

updating the item profile associated with the given one item by writing the received rating into an appropriate one of the item n-tuples or writing the received rating into an appropriate one of the item entries; and

calculating, for the given one user, a plurality of said similarity factors, wherein each of the similarity factors for said given one user represents similarity between said given one user and another one of the users with respect to the items.

8. A computer readable medium having computer executable instructions stored therein, which, when executed by a computer, perform a method for recommending an item to one of a plurality of users, wherein the item is not yet rated by said one user, the method comprising the steps, implemented through said instructions, of:

(a) storing a user profile, in the memory, for each of a plurality of users, wherein the user profile comprises a separate rating value, supplied by a particular one of the users, for each corresponding one of a plurality of items, said items including the item non-rated by the user;

(b) storing an item profile, in the memory, for each of the rated items, wherein the item profile comprises a separate rating value, for a particular one of the items, provided by each one of the plurality of the users, wherein the user profile and the item profile are distinct from each other;

(c) calculating, for each one of the plurality of users and in response to the user and item profiles, a plurality of similarity factors, between said each one user and at least one other one of the users, for each of said items, including said non-rated item;

(d) selecting, in response to the plurality of similarity factors and for each one of the plurality of users, a plurality of neighboring ones of the users, such that each of the neighboring ones of the users has an associated similarity factor which is greater than a first predefined threshold value or, if a confidence factor is associated with the associated similarity factor, both the associated similarity factor is less than the first predefined threshold and the confidence factor exceeds a second predefined threshold value;

(e) assigning a corresponding weight to each of the neighboring users so as to define a plurality of weights; and

(f) recommending at least one of a plurality of the items to said one user in response to the plurality of weights and ratings given to the non-rated item by the neighboring ones of the users.