Recommender system

Abstract

A recommender system may he used to predict a user behavior that a user will give in relation to an item. In an embodiment such predictions are used to enable items to be recommended to users. For example, products may be recommended to customers, potential friends may be recommended to users of a social networking tool, organizations may be recommended to automated users or other items may be recommended to users. In an embodiment a memory stores a data structure specifying a bi-linear collaborative filtering model of user behaviors. In the embodiment an automated inference process may be applied to the data structure in order to predict a user behavior given information about a user and information about an item. For example, the user information comprises user features as well as a unique user identifier.

What is claimed is: 
     
         1 . A computer-implemented method comprising the steps of:
 arranging a memory to store a data structure specifying a rating model comprising an inner product of a user trait vector and an item trait vector, the user trait vector comprising a set of user features including a unique user identifier and the item trait vector comprising a set of item features including a unique item identifier;   storing statistics describing probability distributions associated with the user trait vector and item trait vector;   applying an inference process to the data structure in order to update the statistics on the basis of an observed rating and associated user features and item features; and   receiving user features and item features and predicting a rating using the data structure.   
     
     
         2 . The method as claimed in  claim 1  wherein the data structure is a factor graph. 
     
     
         3 . The method as claimed in  claim 2  wherein the inference process comprises message passing over the factor graph. 
     
     
         4 . The method as claimed in  claim 1  which further comprises recommending an item to a user on the basis of the predicted rating 
     
     
         5 . The method as claimed in  claim 1  which further comprises updating the statistics by adding noise to the probability distributions. 
     
     
         6 . The method as claimed in  claim 1  which further comprises converting the predicted rating into a predicted ranking using threshold values. 
     
     
         7 . The method as claimed in  claim 6  which further comprises learning at least two threshold values for each of a plurality of users. 
     
     
         8 . One or more computer storage media, the one or more computer storage media being hardware storing computer-readable instructions that when executed by a processor causes the processor to perform acts, the acts comprising:
 arranging a memory to store a data structure specifying a rating model comprising an inner product of a user trait vector and an item trait vector, the user trait vector comprising a set of user features including a unique user identifier and the item trait vector comprising a set of item features including a unique item identifier;   storing statistics describing probability distributions associated with the user trait vector and item trait vector;   applying an inference process to the data structure in order to update the statistics on the basis of an observed rating and associated user features and item features; and   receiving user features and item features and predicting a rating using the data structure.   
     
     
         9 . The one or more computer storage media as claimed in  claim 8  wherein the data structure is a factor graph. 
     
     
         10 . The one or more computer storage media as claimed in  claim 9  wherein the inference process comprises message passing over the factor graph. 
     
     
         11 . The one or more computer storage media as claimed in  claim 8  wherein the acts further comprise recommending an item to a user on the basis of the predicted rating. 
     
     
         12 . The one or more computer storage media as claimed in  claim 8  wherein the acts further comprise updating the statistics by adding noise to the probability distributions. 
     
     
         13 . The one or more computer storage media as claimed in  claim 8  wherein the acts further comprise converting the predicted rating into a predicted ranking using threshold values. 
     
     
         14 . The one or more computer storage media as claimed in  claim 13  wherein the acts further comprise learning at least two threshold values for each of a plurality of users. 
     
     
         15 . A recommender system comprising:
 one or more processors operable with one or more memories and configured to:
 arrange a memory to store a data structure specifying a rating model comprising an inner product of a user trait vector and an item trait vector, the user trait vector comprising a set of user features including a unique user identifier and the item trait vector comprising a set of item features including a unique item identifier; 
   store statistics describing probability distributions associated with the user trait vector and item trait vector;   apply an inference process to the data structure in order to update the statistics on the basis of an observed rating and associated user features and item features; and   receive user features and item features and predicting a rating using the data structure.   
     
     
         16 . The recommender system as claimed in  claim 15  wherein the data structure is a factor graph. 
     
     
         17 . The recommender system as claimed in  claim 16  wherein the inference process comprises message passing over the factor graph. 
     
     
         18 . The recommender system as claimed in  claim 15  wherein the one or more processors are further configured to recommend an item to a user on the basis of the predicted rating. 
     
     
         19 . The recommender system as claimed in  claim 15  wherein the one or more processors are further configured to update the statistics by adding noise to the probability distributions. 
     
     
         20 . The recommender system as claimed in  claim 15  wherein the one or more processors are further configured to convert the predicted rating into a predicted ranking using threshold values.