Recommendations system

Patent metadata

Field	Value
Publication number	US-9922360-B2
Application number	US-201313928275-A
Country	US
Kind code	B2
Filing date	Jun 26, 2013
Priority date	Jun 26, 2012
Publication date	Mar 20, 2018
Grant date	Mar 20, 2018

Abstract

A recommendation system communicates with end user devices over a computer communications network and includes a recommendations engine that communicates with the network from behind a firewall that inhibits unauthorized communications from the network, a data file system that receives disposition update data in response to purchase actions initiated by an end user device and receives ingested user activity data in response to activity at a network storefront initiated by an end user device, and a database system that updates a recommendation model for the end user device in response to receiving the disposition update data and ingested user activity data. The recommendations engine generates a recommendation for the end user device in response to the update of the recommendation model.

First claim

What is claimed is:

1. A non-transitory computer-readable storage medium carrying one or more sequences of one or more instructions stored thereon executed by a processor to perform a method of processing communications with at least one end user mobile device over a computer communications network, the method comprising:
receiving user activity data at a front end data center of a recommendations system sent over the network from a virtual network storefront accessed by the at least one end user mobile device, or from a mobile application on the at least one end user mobile device, wherein the user activity data comprises of log files of user clicks and online purchases;
updating an ingestion queue table storage based on the received user activity data;
receiving a polling query at the ingestion queue table storage, wherein the polling query is by a back end processing component that is behind a firewall that inhibits unauthorized communications from the network, and wherein the polling query by the back end processing component determines if there are any new entries of the user activity data in the ingestion queue that were added since a previous polling occurred;
at a time the polling query indicates new entries of the user activity data in the ingestion queue, copying the new entries of the user activity data from the ingestion queue table storage located in a front end processing component of the front end data center to a file storage of the back end processing component in response to the polling query;
deleting the new entries of the user activity data from the ingestion queue table storage after the copying the new entries of the user activity data, wherein the back end processing component stores a recommendation model in a memory storage, and the back end processing component updates the recommendation model for the at least one end user mobile device based on the sent new entries of the user activity data;
receiving, at the back end component, the updated recommendation model;
storing, at the back end component, the updated recommendation model in a database, wherein the database and the memory storage comprise a database system; and
generating a recommendation for the at least one end user mobile device based on the updated recommendation model in response to a recommendation request from the at least one end user mobile device or the virtual network storefront.

2. The non-transitory computer-readable storage medium as in claim 1 , wherein the recommendation request includes a user agent string of the end user mobile device and resolves a device ID that indicates items in the database with which the end user mobile device is compatible.

3. The non-transitory computer-readable storage medium as in claim 2 , wherein the back end processing component computes coefficients for the compatible items in accordance with the recommendations model and the end user mobile device.

4. The non-transitory computer-readable storage medium as in claim 2 , wherein the method further comprises determining a customer identifier that is contained in each received user activity data, received catalog data, and each recommendation request, such that the customer identifier determines a customer specific partition of data in the database system, regardless of a size of the database system.

5. The non-transitory computer-readable storage medium as in claim 2 , wherein the recommendation model is computed such that a database item having a predetermined item type determines a target item in the database system in accordance with the recommendation request as having either the same item type or a different item type.

6. The non-transitory computer-readable storage medium as in claim 2 , wherein the database system includes at least one user profile table and wherein the user activity data is stored into the user profile table for immediate use in generating the recommendation.

7. The non-transitory computer-readable storage medium as in claim 2 , wherein one or more descriptive tags are added to data for items in the database system such that the tags may be utilized by the database system to generate the recommendation model.

8. A method of processing communications with at least one end user mobile device over a computer communications network, the method comprising:
receiving, by a front end data center comprising of one or more storage devices and one or more processors of a recommendation system, user activity data sent over the network from a virtual network storefront accessed by the at least one end user mobile device, or from a mobile application on the at least one end user mobile device, wherein the user activity data comprises of log files of user clicks and online purchases;
updating, by the one or more processors of the recommendations system, an ingestion queue table storage based on the received user activity data;
receiving, by a back end processing component comprising of the one or more processors of the recommendation system, a polling query at the ingestion queue table storage, wherein the back end processing component is behind a firewall that inhibits unauthorized communications from the network;
determining, by the back end processing component, new entries of the user activity data in the ingestion queue table storage have been added since a previous polling;
in response to the determining, copying, by the one or more processors of the recommendations system, the new entries of the user activity data from the ingestion queue table storage of a front end processing component of the front end data center to a file storage of the back end processing component;
deleting the new entries of the user activity data from the ingestion queue table storage after the copying the new entries of the user activity data, wherein the back end processing component stores a recommendation model in a memory storage, and the back end processing component updates the recommendation model for the at least one end user mobile device based on the sent new entries of the user activity data;
receiving, by the back end component, the updated recommendation model;
storing, by the back end component, the updated recommendation model in a database, wherein the database and the memory storage comprise a database system; and
generating, by the one or more processors of the recommendations system, a recommendation for the at least one end user mobile device based on the updated recommendation model in response to a recommendation request from the at least one end user mobile device or the virtual network storefront.

9. The method as in claim 8 , wherein the recommendation request includes a user agent string of the end user mobile device and resolves a device ID that indicates items in the database with which the end user mobile device is compatible.

10. The method as in claim 9 , wherein the back end processing component computes coefficients for the compatible items in accordance with the updated recommendation model and the end user mobile device.

11. The method as in claim 9 , wherein the method further determines a customer identifier that is contained in each received user activity data, received catalog data, and each recommendation request, such that the customer identifier determines a customer specific partition of data in the database, regardless of a size of the database system.

12. The method as in claim 9 , wherein the recommendation model is computed such that a database item having a predetermined item type determines a target item in the database system in accordance with the recommendation request as having either the same item type or a different item type.

Assignees

• Here Global Bv

Inventors

• Hendrick Daniel
• Lanphear Damon
• Mahfoud Robert
• Megraw Rodrick

Classifications

• G06Q30/0601

  Electronic shopping [e-shopping] · CPC title

• G06Q30/02 Primary

  Marketing; Price estimation or determination; Fundraising · CPC title

• G06Q30/0631 Primary

  Recommending goods or services · CPC title