Advertisement target determining device and advertisement target determining method

Patent metadata

Field	Value
Publication number	US-12067593-B2
Application number	US-202318123105-A
Country	US
Kind code	B2
Filing date	Mar 17, 2023
Priority date	Mar 17, 2022
Publication date	Aug 20, 2024
Grant date	Aug 20, 2024

Abstract

Provided is a method of determining an advertisement target according to an advertisement request, the method includes: obtaining usage history information from a plurality of devices, obtaining features of the plurality of devices, based on the usage history information, and generating feature vectors for the obtained features; determining labels for the plurality of devices, based on the advertisement request and the obtained features; generating an advertisement target inference model, based on the determined labels and the feature vectors; and determining at least one advertisement target device among the plurality of devices by applying the generated advertisement target inference model to the plurality of devices.

First claim

What is claimed is:

1. A method of determining an advertisement target according to an advertisement request, the method comprising:
obtaining usage history information from a first plurality of devices via transmission from a communication interface of a device of the first plurality of devices;
obtaining a first plurality of features of the first plurality of devices, based on the usage history information;
generating a plurality of feature vectors for the first plurality of features;
determining a first plurality of labels for the first plurality of devices, based on the advertisement request and the first plurality of features by:
extracting a second plurality of devices that have achieved an advertisement purpose with respect to an advertisement object included in the advertisement request from among the first plurality of devices, based on the usage history information;
determining a second plurality of labels of the second plurality of devices as a first value;
extracting a third plurality of devices from among the first plurality of devices for which labels have not been determined with a third plurality of features that are similar to a second plurality of features the second plurality of devices;
training a neural network model comprising a plurality of lavers and a first plurality of weights by:
providing the neural network model the advertise request and the first plurality of features as a plurality of input values; and
determining a second plurality of weights by optimizing a cost value of the first plurality of weights;
outputting the second plurality of weights of the first plurality of features;

determining one or more third labels of the third plurality of devices as the first value based on the third plurality of features and the second plurality of weights; and
determining one or more fourth labels of the third plurality of devices for which the labels have not been determined as a second value;

generating an advertisement target inference model, based on the first plurality of labels and the plurality of feature vectors; and
determining at least one advertisement target device among the first plurality of devices by applying the generated advertisement target inference model to the first plurality of devices.

2. The method of claim 1 , wherein the neural network model is trained using a plurality of priorities of the first plurality of features, and the plurality of priorities are determined through experimentation.

3. The method of claim 1 , wherein the first plurality of features comprise at least one of information about a device user including age, gender, region and income information, advertisement viewing information, advertisement click information, application usage information, external device usage information, television (TV) viewing information, video on demand (VOD) viewing information, application installation and removal information, device usage pattern information, or played game information.

4. The method of claim 1 , wherein the advertisement request comprises an advertisement sector, the advertisement object, and the advertisement purpose.

5. The method of claim 1 , wherein the generating of the advertisement target inference model comprises generating the advertisement target inference model by using a neural network trained to generate the advertisement target inference model from the first plurality of labels and the plurality of feature vectors.

6. The method of claim 1 , wherein the determining of the at least one advertisement target device comprises:
obtaining a score by applying the generated advertisement target inference model to the first plurality of devices; and
determining, as the advertisement target, at least one device with the score equal to or greater than a preset value among the first plurality of devices, based on the obtained score.

7. The method of claim 6 , further comprising determining the at least one advertisement target device by filtering the determined at least one advertisement target device, based on an advertisement condition that is included in the advertisement request.

8. The method of claim 1 , wherein the obtaining of the usage history information from the first plurality of devices comprises refining the usage history information by applying a square root or logarithm.

9. An advertisement target determining device for determining an advertisement target according to an advertisement request, the advertisement target determining device comprising:
a memory that stores one or more instructions; and
a processor configured to execute the one or more instructions that are stored in the memory, to cause the processor to:
obtain usage history information from a first plurality of devices via transmission from a communication interface of a device of the first plurality of devices;
obtain a first plurality of features of the first plurality of devices, based on the usage history information;
generate a plurality of feature vectors for the first plurality of features;
determine a first plurality of labels for the first plurality of devices, based on the advertisement request and the first plurality of features by:
extracting a second plurality of devices that have achieved an advertisement purpose with respect to an advertisement object included in the advertisement request from among the first plurality of devices, based on the usage history information;
determining a second plurality of labels of the second plurality of devices as a first value;
extracting a third plurality of devices from among the first plurality of devices for which labels have not been determined with a third plurality of features that are similar to a second plurality of features the second plurality of devices;
training a neural network model comprising a plurality of lavers and a first plurality of weights by:
providing the neural network model the advertisement request and the first plurality of features as a plurality of input values; and
determining a second plurality of weights by optimizing a cost value of the first plurality of weights;
outputting the second plurality of weights of the first plurality of features;

determining one or more third labels of the third plurality of devices as the first value based on the third plurality of features and the second plurality of weights; and
determining one or more fourth labels of the third plurality of devices for which the labels have not been determined as a second value;

generate an advertisement target inference model, based on the first plurality of labels and the plurality of feature vectors; and
determine at least one advertisement target device among the first plurality of devices by applying the generated advertisement target inference model to the first plurality of devices.

10. The advertisement target determining device of claim 9 , wherein the neural network model is trained using a plurality of priorities of the first plurality of features, and the plurality of priorities are determined through experimentation.

11. The advertisement target determining device of claim 9 , wherein the first plurality of features comprise at least one of information about a device user including age, gender, region and income information, advertisement viewing information, advertisement click information, application usage information, external device usage information, television (TV) viewing information, video on demand (VOD) viewing information, application installation and removal information, device usage pattern information, or played game information.

12. The advertisement target determining device of claim 9 , wherein the advertisement

Assignees

• Samsung Electronics Co Ltd

Inventors

• Park Sanghun
• Jo Kunhee
• Joe Seongmin
• Chi Yeongjin

Classifications

• G06N5/04

  Inference or reasoning models · CPC title

• H04N21/44222

  Analytics of user selections, e.g. selection of programmes or purchase activity (monitoring of user selections in data processing systems G06F11/34; arrangements for monitoring the user's behaviour or opinions in broadcast systems H04H60/33) · CPC title

• H04N21/2542

  for selling goods, e.g. TV shopping (payment schemes, payment architectures or payment protocols for electronic shopping systems G06Q20/12) · CPC title

• H04N21/25841

  involving the geographical location of the client (retrieval from the Internet by querying based on geographical locations G06F16/9537; arrangements for identifying locations of receiving stations in broadcast systems H04H60/51; location of the user terminal in data switching networks H04L67/52; services making use of the location of users or terminals in wireless networks H04W4/02; locating users or terminals in wireless networks H04W64/00) · CPC title

• G06N20/00

  Machine learning · CPC title