Probabilistic clustering of an item

Patent metadata

Field	Value
Publication number	US-9852193-B2
Application number	US-69488510-A
Country	US
Kind code	B2
Filing date	Jan 27, 2010
Priority date	Aug 10, 2009
Publication date	Dec 26, 2017
Grant date	Dec 26, 2017

Abstract

A clustering and recommendation machine determines that an item is included in a cluster of items. The machine accesses item data descriptive of the item. The machine accesses a vector that represents the cluster and calculates the likelihood that the item is included in the cluster, based on the item variable and the probability parameter. The machine determines that the item is included in the cluster, based on the likelihood. The machine also recommends an item to a potential buyer. The machine accesses behavior data that represents a first event type pertinent to a first cluster of items. The machine calculates a probability that a second event type pertaining to a second cluster of items will co-occur with the first event type. The machine identifies an item from the second cluster to be recommended and presents a recommendation of the item to the potential buyer.

First claim

What is claimed is:

1. A computer-implemented method comprising:
accessing item data that corresponds to an item available for sale and inclusive of an item variable that represents information describing an attribute of the item, the item variable being a binary variable that represents an occurrence of a term in textual data descriptive of the item available for sale;
accessing a vector that is representative of a cluster of items having attributes corresponding to the attribute of the item, the vector representing a product of which the item is a specimen, the vector including a probability distribution that is pertinent to the cluster of items and represented as a plurality of probability parameters that include a probability parameter modeling the binary variable that represents the occurrence of the term in the textual data descriptive of the item available for sale;
calculating a result based on the binary variable that represents the occurrence of the term in the textual data descriptive of the item available for sale including calculating a logarithm based on the probability parameter modeling the binary variable and calculating the result based on the logarithm, the result representing a likelihood that the item is a specimen of the product represented by the cluster of items having attributes corresponding to the attribute of the item and represented by the vector, the calculating being performed by a module implemented using a processor of a machine;
determining that the item is included in the cluster based on the result; and
presenting a recommendation of the item available for sale to a potential buyer of the item, the recommendation indicating that the item is the specimen of the product represented by the cluster of items.

2. The computer-implemented method of claim 1 , further comprising accessing item data that includes at least one of:
a categorical variable representing an attribute pertinent to the item;
or a continuous variable representing a number pertinent to the item.

3. The computer-implemented method of claim 1 , wherein the probability distribution is selected from a group consisting of:
a binomial distribution;
a multinomial distribution; and
a Gaussian distribution.

4. The computer-implemented method of claim 1 , wherein the probability parameter is selected from a group consisting of:
a Bernoulli success probability;
a multinomial parameter; and
a Gaussian mean.

5. The computer-implemented method of claim 1 , wherein the calculating of the result includes calculating a sum based on the logarithm and based on the item variable.

6. The computer-implemented method of claim 1 , wherein the calculating of the result includes calculating a sum based on the logarithm and not based on the item variable.

7. The computer-implemented method of claim 1 further comprising:
calculating a sum based on the result, the sum representing a total logarithmic likelihood that a plurality of items is included in the cluster;
calculating an argument of a maximum of the sum; and
estimating the vector representative of the cluster based on the argument of the maximum.

8. The computer-implemented method of claim 1 , wherein the item data is generated by a seller of the item, the method further comprising receiving the item data.

9. The computer-implemented method of claim 1 , wherein the item data includes at least one of a title, a description, an attribute name, an attribute value, a price, or a size of the item.

10. The computer-implemented method of claim 1 further comprising:
storing a map file that includes a correspondence between the item and the cluster; and
performing a lookup operation using the map file in response to a query pertinent to at least one of the item or the product represented by the cluster of items.

11. The computer-implemented method of claim 1 further comprising:
determining a size of the cluster of items, the size indicating a number of specimens of the product represented by the cluster;
performing a comparison of the size to a threshold; and
determining that the cluster is to be discarded based on the comparison.

12. A system comprising:
an access module, implemented using at least one hardware processor, configured to:
access item data that corresponds to an item available for sale and inclusive of an item variable that represents information describing an attribute of the item, the item variable being a binary variable that represents an occurrence of a term in textual data descriptive of the item available for sale; and
access a vector that is representative of a cluster of items having attributes corresponding to the attribute of the item, the vector representing a product of which the item is a specimen, the vector including a probability distribution that is pertinent to the cluster of items and represented as a plurality of probability parameters that include a probability parameter modeling the binary variable that represents the occurrence of the term in the textual data descriptive of the item available for sale;

a processor configured by a probability module that configures the processor to calculate a result based on the binary variable that represents the occurrence of the term in the textual data descriptive of the item available for sale including calculating a logarithm based on the probability parameter modeling the binary variable and calculating the result based on the logarithm, the result representing a likelihood that the item is a specimen of the product represented by the cluster of items having attributes corresponding to the attribute of the item and represented by the vector;
a determination module to determine that the item is included in the cluster based on the result; and
a recommendation module to present a recommendation of the item available for sale to a potential buyer of the item, the recommendation indicating that the item is the specimen of the product represented by the cluster of items.

13. The system of claim 12 , wherein the probability module, in calculating the result, is to calculate a sum based on the logarithm and not based on the item variable.

14. The system of claim 12 further comprising an estimation module to calculate a sum based on the result, the sum representing a total logarithmic likelihood that a plurality of items is included in the cluster;
calculate an argument of a maximum of the sum; and
estimate the vector representative of the cluster based on the argument of the maximum.

15. A non-transitory machine-readable storage medium comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising:
accessing item data that corresponds to an item available for sale and inclusive of an item variable that represents information describing an attribute of the item, the item variable being a binary variable that represents an occurrence of a term in textual data descriptive of the item available for sale;
accessing a vector that is representative of a cluster of items having attributes corresponding to the attribute of the item, the vector representing a product of which the item is a specimen, the vector including a probability distribution that is pertinent to the cluster of items and represented as a plurality of probability parameters that include a probability parameter modeling the binary variable that represents the occurrence of the term in the textual data descriptive of the item available for sale;
calculating a result based on the binary variable that represents the occurrence of the term in the textual d

Assignees

• Chen Ye
• Canny John
• Ebay Inc

Inventors

• Chen Ye
• Canny John

Classifications

• G06F18/2321

  using statistics or function optimisation, e.g. modelling of probability density functions · CPC title

• G06F16/2462 Primary

  Approximate or statistical queries · CPC title

• G06F17/30536 Primary

  Physics · mapped topic

• G06K9/6226

  Physics · mapped topic