EP0751471A1
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Method and apparatus for item recommendation using automated collaborative filtering
Assignee: Massachusetts Institute of Technology
•
Filing: 1996-06-18
•
Granted: —
•
Expires: 2016-06-18
Abstract
Abstract A method for recommending items to users using automated collaborative filtering stores profiles of users relating ratings to items in memory. Profiles of items are also stored in memory, the item profiles associating users with the rating given to the item by that user. Similarity factors with respect to other users are calculated for a user, and these similarity factors are used to select a set of neighboring users. The neighboring users are weighted based on their respective similarity factors, and a rating for an item contained in the domain is predicted. In one embodiment, items in the domain have features. In this embodiment, the values for features can be clustered, and the similarity factors incorporate assigned feature weights and feature value cluster weights.
Inventors
- 👤 Yezdezard Z. Lashkari
- 👤 Patricia Maes
- 👤 Max E. Metral
- 👤 Upendra Shardanand
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-11-05 08:16
CPC Classifications
H; H04; H04N; H04N21/00; H04N21/20; H04N21/25; H04N21/251; H04N21/252; G; G06; G06F; G06F16/00; G06F16/30; G06F16/33; G06F16/335; G06F16/90; G06F16/95; G06F16/953; G06F16/9535; G06Q; G06Q30/00; G06Q30/02; G06Q30/06
Claims Analysis
Total Claims: 21
Claim 1
Independent Claim
250 words
A method for recommending an item to one of a plurality of users, the item not yet rated by the user, the method comprising the steps of: (a) storing a user profile in a memory for each of a plurality of users, wherein the user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; (b) storing an item profile in a memory for each of the plurality of items, each of the plurality of items belonging to one of a plurality of groups, wherein the item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; (c) calculating, for each of the plurality of users, a plurality of similarity factors, each of the plurality of similarity factors representing the similarity between each user and another of the plurality of users based on item ratings for a particular group; (d) selecting, for each of the plurality of users, a plurality of neighboring users with respect to each group, the selection responsive to the similarity factors; (e) assigning a weight to each of the neighboring users; and (f) recommending an item to one of the plurality of users based on the weights assigned to the user's neighboring users and the ratings given to the unrated item by the user's neighboring users.
Claim 2
Dependent Claim
107 words
The method of claim 1 wherein step (c) further comprises: (c-a) retrieving the item profile for one of the plurality of items that has been rated by one of the plurality of users; (c-b) determining, from the item's profile, other users that have also rated the item; (c-c) calculating a similarity factor between the one user and each of the plurality of other users that have also rated the item; (c-d) repeating steps (c-a) through (c-c) until all items rated by the one user for the one group have been retrieved; and (c-e) repeating steps (c-a) though (c-d) until similarity factors for each user have been calculated.
Claim 3
Dependent Claim
47 words
The method of claim 2 wherein step (c-c) further comprises calculating a similarity factor between the one user and each of the plurality of other users that have also rated the item, said similarity factor based only on ratings for other items belonging to the same group.
Claim 4
Independent Claim
40 words
The method of any preceding claim wherein step (d) comprises selecting, for each of the plurality of users, a plurality of neighboring users in each group, each selected neighboring user having a similarity factor less than a predetermined threshold value.
Claim 5
Dependent Claim
36 words
The method of claim 4 wherein step (e) further comprises: subtracting, for each neighboring user, the similarity factor for that neighboring user from a predetermined threshold value and dividing each difference by the predetermined threshold value.
Claim 6
Independent Claim
72 words
The method of any preceding claim wherein step (f) comprises: predicting a rating for each item in one of the plurality of groups not yet rated by one of the plurality of users by taking a weighted average of the ratings given to the items in the group by the one user's neighboring users; and recommending a predetermined number of items from the group based on the predicted ratings for those items.
Claim 7
Independent Claim
54 words
The method of any of claims 1 to 5 wherein step (f) comprises: receiving an item selection from one of the plurality of users; and predicting a rating for the selected item by taking a weighted average of the ratings given to the selected item by the one user's neighboring users for the group.
Claim 8
Independent Claim
75 words
The method of any of claims 1 to 5 wherein step (f) comprises: receiving a group selection from one of the plurality of users; predicting a rating for items in the selected group by taking a weighted average of the ratings given to the items in the group by the one user's neighboring users for that group; and recommending a predetermined number of items in the group based on the predicted ratings for those items.
Claim 9
Independent Claim
20 words
The method of any preceding claim further comprising the step of: (g) displaying information about recommended items on a display.
Claim 10
Independent Claim
13 words
The method of any preceding claim wherein an item belongs to multiple groups.
Claim 11
Independent Claim
194 words
A method for recommending, to one of a plurality of users, other users, the method comprising the steps of: (a) storing a user profile in a memory for each of a plurality of users, wherein the user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; (b) storing an item profile in a memory for each of the plurality of items, each of the plurality of items belonging to one of a plurality of groups, wherein the item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; (c) calculating, for each of the plurality of users, a plurality of similarity factors, each of the plurality of similarity factors representing the similarity between each user and another one of the plurality of users based on the item ratings for a particular group; (d) recommending at least one of the neighboring users to one of the plurality of users based on the similarity factors.
Claim 12
Dependent Claim
48 words
The method of claim 11 wherein step (d) further comprises recommending at least one of the neighboring users to one of the plurality of users based on the similarity factors and the number of items rated by both the one user and the at least one neighboring user.
Claim 13
Independent Claim
267 words
A method for predicting a rating for an item, the item having at least one feature defining a characteristic of the item and having a plurality of possible values that may be grouped into clusters of feature values, the method comprising the steps of: (a) storing a user profile in a memory for each of a plurality of users, wherein the user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; (b) storing an item profile in a memory for each of the plurality of items, wherein the item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; (c) assigning, for each user, a weight to each value cluster within each feature; (d) assigning, for each user, a weight to each feature; (e) calculating, for each of the plurality of users, a plurality of similarity factors, each of the plurality of similarity factors based on the feature weights, the cluster weights, and the ratings given to items by the respective users; (f) selecting, for each of the plurality of users, a plurality of neighboring users responsive to the similarity factors; (g) assigning a weight to each of the neighboring users; and (h) recommending an item to one of the plurality of users based on the weights assigned to the user's neighboring users and the ratings given to the unrated item by the user's neighboring users.
Claim 14
Dependent Claim
42 words
The method of claim 13 wherein step (c) further comprises: assigning, for each user, a weight to each value cluster within each feature based on the rating given to the item by the user and the number of feature value clusters present.
Claim 15
Dependent Claim
31 words
The method of claim 13 or 14 wherein step (d) further comprises: assigning, for each user, a weight to each feature based on the number of features defined for the item.
Claim 16
Dependent Claim
31 words
The method of claim 13 or 14 wherein step (d) further comprises: assigning, for each user, a weight to each feature based on the weights assigned to each feature value cluster.
Claim 17
Dependent Claim
27 words
The method of claim 16, wherein the feature weight is assigned by dividing the standard deviation of the cluster weights by the mean of the cluster weights.
Claim 18
Independent Claim
47 words
The method of any of claims 13 to 17, wherein step (c) further comprises the steps of: (c-a) calculating, for each feature value, a similarity factor to all other feature values in a feature; and (c-b) grouping the feature values into clusters based on the similarity factors.
Claim 19
Independent Claim
251 words
A method for recommending an item to one of a plurality of users, the item having at least one feature defining a characteristic of the item and having a plurality of possible values that may be grouped into clusters of feature values, the method comprising the steps of: (a) storing a user profile in a memory for each of a plurality of users, wherein the user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; (b) storing an item profile in a memory for each of the plurality of items, wherein the item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; (c) assigning, for at least one user, a weight to each value cluster within each feature; (d) assigning, for the at least one user, a weight to each feature; (e) calculating, for each of the plurality of items, a plurality of similarity factors, each of the plurality of similarity factors based on the feature weights and the cluster weights; (f) selecting an item for which a favourable rating has been received from the at least one user; (g) selecting a plurality of items responsive to the selected item and the similarity factors; (h) recommending at least one of the selected items to the at least one user.
Claim 20
Dependent Claim
300 words
An article of manufacture having program means embodied therein, the program means for predicting a rating for an item, the item having at least one feature defining a characteristic of the item and having a plurality of possible values that may be grouped into clusters of feature values, the article of manufacture comprising: computer-readable program means for storing a user profile in a memory for each of a plurality of users, wherein the user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; computer-readable program means for storing an item profile in a memory for each of the plurality of items, wherein the item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; computer-readable program means for assigning, for each user, a weight to each value cluster within each feature; computer-readable program means for assigning, for each user, a weight to each feature; computer-readable program means for calculating, for each of the plurality of users, a plurality of similarity factors, each of the plurality of similarity factors based on the feature weights, the cluster weights, and the ratings given to items by the respective users; computer-readable program means for selecting, for each of the plurality of users, a plurality of neighboring users responsive to the similarity factors; computer-readable program means for assigning a weight to each of the neighboring users; and computer-readable program means for recommending an item to one of the plurality of users based on the weights assigned to the user's neighboring users and the ratings given to the unrated item by the user's neighboring users.
Claim 21
Independent Claim
249 words
An apparatus for recommending an item, the item having at least one feature defining a characteristic of the item and having a plurality of possible values that may be grouped into clusters of feature values, comprising: a memory element for storing user profiles, wherein each user profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to one of a plurality of items by the user; a memory element for storing item profiles, wherein each item profile includes a plurality of values, each of at least some of the plurality of values representing a rating given to the item by one of the plurality of users; means for assigning, for each user, a weight to each value cluster within each feature; means for assigning, for each user, a weight to each feature; means for calculating, for each of the plurality of users, a plurality of similarity factors, each of the plurality of similarity factors based on the feature weights, the cluster weights, and the ratings given to items by the respective users; means for selecting, for each of the plurality of users, a plurality of neighboring users responsive to the similarity factors; means for assigning a weight to each of the neighboring users; and means for recommending an item to one of the plurality of users based on the weights assigned to the user's neighboring users and the ratings given to the unrated item by the user's neighboring users.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.