Systems and methods for predicting sales of item listings

Patent metadata

Field	Value
Publication number	US-9727616-B2
Application number	US-49820909-A
Country	US
Kind code	B2
Filing date	Jul 6, 2009
Priority date	Jul 6, 2009
Publication date	Aug 8, 2017
Grant date	Aug 8, 2017

Abstract

A method and a system are disclosed for predicting sales of item listings on a network-based system. For example, historical transaction data generated by the network-based system is accessed to create a prediction model. A feature predictive of an item being sold through the network-based system is selected. A training set is created by extracting the predictive feature from the historical transaction data. The prediction model is trained based on the training data set to predict the probability of an item listing being sold through the network-based system. The prediction model is used to rank search results, and the search results can be presented through the network-based system.

First claim

What is claimed is:

1. A method to rank search results comprising:
accessing, using one or more processors within a network-based system, historical transaction data;
selecting, using the one or more processors, a plurality of predictive features that are predictive of an item being sold over the network-based system, selecting the plurality of predictive features includes using a regression model to determine the plurality of predictive features with the highest correlation to items being sold, wherein the selecting includes selecting a category related feature, a seller related feature, and an item related feature;
creating, using the one or more processors, a training data set including extracting the plurality of predictive features from the historical transaction data;
creating, using the one or more processors, a prediction model based on the training data set to predict a probability that an item offered for sale by an item listing published on the network-based system will be sold;
receiving a search request on the network-based system;
generating, using the one or more processors, a list of search results based on the search request;
ranking, using the one or more processors, the list of search results using the prediction model; and
presenting, using the one or more processors, the ranked search results.

2. The method of claim 1 , wherein the creating of the prediction model includes learning a regression model from the training data set.

3. The method of claim 2 , wherein the learning of the regression model includes using at least one of linear regression and logistic regression techniques.

4. The method of claim 1 , wherein the plurality of predictive features further includes at least one of:
item velocity;
item popularity;
number of competitors on the network-based system;
conversion rate of similar items;
price;
sale type;
item quantity;
item return policy level;
best offer, and
store inventory.

5. The method of claim 1 , wherein the creating of the training data set includes extracting a seller feature from the historical transaction data.

6. The method of claim 5 , wherein the extracting the seller feature includes selecting the seller feature from a group including:
a feedback score;
a percentage of positive feedback;
a seller level; and
a type of seller.

7. A method to rank item listings according to predicted sales in a network-based system, the method comprising:
accessing, using one or more processors within the network-based system, a training data set including historical transaction data generated by the network-based system;
extracting, using the one or more processors, a plurality of item features from the training data set, extracting the plurality of item features includes using a regression model to determine the plurality of item features with the highest correlation to items being sold;
extracting, using the one or more processors, a seller feature from the training data set, the seller feature being an attribute of a seller within the network-based system, the extracting the seller feature includes using a regression model to determine the seller feature with the highest correlation to items being sold;
extracting, using the one or more processors, a category feature from the training data set, the extracting the category feature includes using a regression model to determine the category feature with the highest correlation to items being sold;
creating, using the one or more processors a prediction model using the plurality of item features, the category feature, and the seller feature;
receiving, using the one or more processors, a search request for item listings on the network-based system;
using the prediction model, ranking, using the one or more processors, a result set generated using the search request; and
presenting, using the one or more processors, the ranked search results.

8. The method of claim 7 , wherein the creating of the prediction model includes creating a regression model based on the training data set.

9. The method of claim 8 , wherein the creating of the regression model includes using at least one of linear regression and logistic regression techniques.

10. The method of claim 7 , wherein the plurality of item features further include at least one of:
item velocity;
item popularity;
number of competitors on the network-based system;
conversion rate of similar items;
price;
sale type;
item quantity;
item return policy level;
best offer, and
store inventory.

11. A network-based system to rank search results, the system comprising:
a database to maintain historical transaction data;
a server including a processor coupled to a memory device, the processor and the memory device configured to execute modules and engines including:
a training module to create a training data set including a plurality of predictive features, the plurality of predictive features being predicative of an item offered for sale on the network at the system being successfully sold, the plurality of predictive features being extracted from the historical transaction data maintained in the database, and wherein the training module is to select the plurality of predictive features using a regression model to determine the plurality of predictive features with the highest correlation to items being sold, wherein the plurality of predictive features selected include a category related feature, a seller related feature, and an item related feature;
a prediction module to rank search results based upon a prediction model calculated using the training data set,
a search engine to receive a search request, generate a list of search results based on the search request, and communicate the list of search results to the prediction module for ranking; and
a presentation engine to present the list of search results ranked by the prediction module.

12. The network-based system of claim 11 , wherein the prediction module is to create the prediction model by calculating a regression model from the training data set.

13. The network-based system of claim 12 , wherein the prediction module is to calculate the regression model using at least one of linear regression and logistic regression techniques.

14. The network-based system of claim 11 , wherein the training module is to select the plurality of predictive features from a group including:
category popularity;
item velocity;
item popularity
number of competitors on the network-based system;
conversion rate of similar items;
price;
sale type;
item quantity;
item return policy level;
best offer, and
store inventory.

15. The network-based system of claim 11 , wherein the training module is to create the training data set to further include a seller feature extracted from the historical transaction data.

16. The network-based system of claim 15 , wherein the training module is to select the seller feature from a group including:
a feedback score;
a percentage of positive feedback;
a seller level; and
a type of seller.

17. A non-transitory machine-readable storage medium embodying instructions which, when executed by a computer-implemented network-based system, cause the computer-implemented network-based system to execute a method comprising:
accessing historical transaction data;
selecting a plurality of predictive features that are predictive of an item being sold over the network-based system, selecting the plurality of predictive features includes using a regression model to determine the plurality of predictive

Assignees

• Wu Xiaoyuan
• Bolivar Alvaro
• Paypal Inc

Inventors

• Wu Xiaoyuan
• Bolivar Alvaro

Classifications

• G06F16/24578 Primary

  using ranking · CPC title

• G06Q30/00

  Commerce · CPC title

• G06Q30/0631

  Recommending goods or services · CPC title

• G06Q10/06375

  Prediction of business process outcome or impact based on a proposed change · CPC title

• G06F17/3053 Primary

  Physics · mapped topic