Systems and methods for predicting user segments in real-time

Patent metadata

Field	Value
Publication number	US-11416893-B2
Application number	US-202016866250-A
Country	US
Kind code	B2
Filing date	May 4, 2020
Priority date	Sep 23, 2016
Publication date	Aug 16, 2022
Grant date	Aug 16, 2022

Abstract

Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform recording one or more actions of a user during an online browsing session; predicting, in real-time, a first user attribute for the user from one or more user attributes during the online browsing session based on the one or more actions of the user during the online browsing session; correlating the first user attribute, as predicted, for the user, wherein a first user preference of one or more user preferences is associated with the first user attribute, as predicted; and mapping each respective user preference of the one or more user preferences to a vector space, wherein: two or more user preferences of the one or more user preferences determined to be more similar to one another are closer together on the vector space than two or more user preferences of the one or more user preferences determined to be less similar to one another. Other embodiments are disclosed herein.

First claim

What is claimed is:

1. A system comprising:
one or more processors; and
one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and cause the one or more processors to perform functions comprising:
recording one or more actions of a user during an online browsing session;
predicting, in real-time, a first user attribute for the user from one or more user attributes during the online browsing session based on the one or more actions of the user during the online browsing session;
correlating the first user attribute, as predicted, for the user with a first user preference of one or more user preferences, wherein the first user preference of the one or more user preferences is associated with the first user attribute, as predicted; and
mapping each respective user preference of the one or more user preferences to a vector space, wherein:
two or more user preferences of the one or more user preferences determined to be more similar to one another are closer together on the vector space than two or more user preferences of the one or more user preferences determined to be less similar to one another.

2. The system of claim 1 , wherein the computing instructions are further configured to run on the one or more processors and cause the one or more processors to perform further functions comprising:
coordinating displaying at least two graphical user interface (GUI) elements on a subsequent GUI during the online browsing session, wherein:
a first GUI element of the at least two GUI elements is associated with the first user preference, as correlated;
a second GUI element of the at least two GUI elements is associated with a second user preference of the one or more user preferences; and
the first user preference, as correlated, is proximate to the second user preference in the vector space.

3. The system of claim 1 , wherein the computing instructions are further configured to run on the one or more processors and cause the one or more processors to perform further functions comprising:
determining the first user preference using one or more training webpages mapped to the vector space.

4. The system of claim 1 , wherein:
predicting, in real-time, the first user attribute for the user during the online browsing session comprises:
predicting, in real-time, a respective probability that an online browsing trail of the online browsing session pertains to each respective user attribute of the one or more user attributes; and

the first user attribute for the user comprises:
a highest probability that the online browsing trail pertains to the first user attribute of the one or more user attributes, the highest probability being relative to respective probabilities of other user attributes of the one or more user attributes.

5. The system of claim 4 , wherein predicting, in real-time, the respective probability that the online browsing trail pertains to each respective user attribute of the one or more user attributes comprises:
predicting, in real-time, a respective probability (P) using a formula comprising:
log P(segment=k|Θ, browseTrail)=Σ t z log P(i t , i t+1 |Θ k ), where:
the respective probability (P) is the respective probability that the online browsing trail pertains to each respective user attribute of the one or more user attributes;
k is an attribute of the one or more user attributes;
Θ is a learned model;
t is the online browsing trail;
i t is a product of a previous webpage in the online browsing trail at a website;
i t+1 is a product of a current webpage in the online browsing trail at the website; and
z is a number of pairs of i t and i t+1 being evaluated.

6. The system of claim 4 , wherein the respective probability is normalized by a duration of the online browsing session.

7. The system of claim 4 , wherein the respective probability is weighted by previous webpages in the online browsing trail.

8. The system of claim 1 , wherein predicting, in real-time, the first user attribute for the user during the online browsing session comprises:
predicting, in real-time, the first user attribute for the user during the online browsing session further based on a proximity of webpages accessed during the online browsing session.

9. The system of claim 1 , wherein the first user attribute comprises a demographic of the user or a purpose of the online browsing session.

10. The system of claim 9 , wherein the purpose of the online browsing session comprises at least one of:
browsing for a different person of a particular age;
browsing for the different person of a particular gender;
browsing for home delivery;
browsing for in-store pickup; or
browsing to order an item during the online browsing session.

11. A method implemented via execution of computing instructions configured to run at one or more processors and configured to be stored at non-transitory computer-readable media, the method comprising: recording one or more actions of a user during an online browsing session;
predicting, in real-time, a first user attribute for the user from one or more user attributes during the online browsing session based on the one or more actions of the user during the online browsing session;
correlating the first user attribute, as predicted, for the user with a first user preference of one or more user preferences, wherein the first user preference of the one or more user preferences is associated with the first user attribute, as predicted; and
mapping each respective user preference of the one or more user preferences to a vector space, wherein:
two or more user preferences of the one or more user preferences determined to be more similar to one another are closer together on the vector space than two or more user preferences of the one or more user preferences determined to be less similar to one another.

12. The method of claim 11 , further comprising:
coordinating displaying at least two graphical user interface (GUI) elements on a subsequent GUI during the online browsing session, wherein:
a first GUI element of the at least two GUI elements is associated with the first user preference, as correlated;
a second GUI element of the at least two GUI elements is associated with a second user preference of the one or more user preferences; and
the first user preference, as recorded, is proximate to the second user preference in the vector space.

13. The method of claim 11 further comprising:
determining the first user preference using one or more training webpages mapped to the vector space.

14. The method of claim 11 , wherein:
predicting, in real-time, the first user attribute for the user during the online browsing session comprises:
predicting, in real-time, a respective probability that an online browsing trail of the online browsing session pertains to each respective user attribute of the one or more user attributes; and

the first user attribute for the user comprises:
a highest probability that the online browsing trail pertains to the first user attribute of the one or more user attributes, the highest probability being relative to respective probabilities of other user attributes of the one or more user attributes.

15. The method of claim 14 , wherein predicting, in real-time, the respective probability that the online browsing trail pertains to each respective user attribute of the one or more user attributes comprises:
predicting, in real-time, a respective probability (P) using a formula comprising:
log P(segment=k|Θ, browseTrail)=Σ t z log P(i t , i t+1 |Θ k

Assignees

• Walmart Apollo Llc

Inventors

• Ali Afroza
• Achan Kannan
• Rapaka Venkata Syam Prakash

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06Q30/0255 Primary

  based on user history · CPC title

• H04L67/535

  Tracking the activity of the user (network monitoring arrangements H04L43/00; recording of computer activity G06F11/34) · CPC title

• G06Q30/0251

  Targeted advertisements · CPC title

• H04L67/306

  User profiles · CPC title