# US 11,308,523 B2 · Validating a target audience using a combination of classification algorithms

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,308,523 B2
- Document ID: US11308523B2
- Status: Granted
- Kind: B2
- Application number: 15/457,882
- Filed: 2017-03-13
- Published: 2018-09-13
- Granted: 2022-04-19
- Earliest claimed: 2017-03-13
- Assignees: Adobe Inc.
- Applicants: Adobe Inc.
- Inventors: Wreetabrata Kar; Viswanathan Swaminathan; Sarathkrishna Swaminathan
- Web page: https://patentgrep.com/patent/US11308523B2

## CPC

- G06Q 30/0271: {Personalized advertisement}
- G06Q 30/0254: {based on statistics}
- G06Q 30/0255: {based on user history}
- G06F 16/435: Filtering based on additional data, e.g. user or group profiles
- G06F 16/635: Filtering based on additional data, e.g. user or group profiles
- H04N 21/25883: {being end-user demographical data, e.g. age, family status or address (arrangements for identifying locations of users in broadcast systems H04H60/52)}
- H04N 21/812: {involving advertisement data (advertising per se G06Q30/02)}
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

This disclosure generally covers systems and methods that determine demographic labels for a user or a group of users by using digital inputs within a predictive model for demographic classification. In particular, the disclosed systems and methods use a unique combination of classification algorithms to determine demographic labels for users as a potential audience of digital content items. When applying the combination of classification algorithms, the disclosed systems and methods use a first classification algorithm to determine user-level-latent features for each user within a group of users based on demographic-label statistics associated with particular digital content items. The disclosed systems and methods then use the user-level-latent features and session-level features (from sessions of each user consuming the digital content items) as inputs in a second classification algorithm to determine a demographic label for each user within the group of users.

## Claims

### Claim 1

In a digital medium environment, a computer-implemented method for determining demographic labels using digital inputs within a predictive model for demographic classification comprising:
receiving demographic-label statistics associated with digital content items;
accessing session-level data associated with the digital content items, wherein the session-level data comprises session-level features for sessions in which a user of a plurality of users consumed one or more of the digital content items;
utilizing a first classification algorithm to determine a user-level-latent feature for the user by analyzing subsets of the demographic-label statistics particular to different digital content items consumed by the user and a listing of the different digital content items consumed by the user;
generating a first coefficient learned by a supervised learning model for user-level-latent features processed by the supervised learning model;
generating a second coefficient learned by the supervised learning model for session-level features processed by the supervised learning model; and
utilizing a second classification algorithm to determine a demographic label for the user by analyzing the user-level-latent feature for the user, the session-level features, the first coefficient learned by the supervised learning model, and the second coefficient learned by the supervised learning model.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein utilizing the first classification algorithm comprises:
identifying, from within the demographic-label statistics and for each digital content item within the listing of the different digital content items, a collective probability of a possible demographic label for consumers of each digital content item within the listing of the different digital content items; and
based on the collective probability of the possible demographic label for the consumers, determining a prior probability of the possible demographic label for the user, wherein the user-level-latent feature for the user comprises the prior probability of the possible demographic label for the user.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, wherein determining the prior probability of the possible demographic label for the user comprises determining a product of the collective probability of the possible demographic label for the consumers of each digital content item within the listing of the different digital content items.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein utilizing the first classification algorithm comprises:
assigning one classifier of a plurality of classifiers to each digital content item within the listing of the different digital content items; and
based on the assigned one classifier of the plurality of classifiers to each digital content item within the listing of the different digital content items, generating a count of the different digital content items within the listing of the different digital content items assigned to each classifier of the plurality of classifiers, wherein the user-level-latent feature for the user comprises the count of the different digital content items within the listing of the different digital content items assigned to each classifier of the plurality of classifiers.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, wherein assigning one classifier of the plurality of classifiers to each digital content item within the listing of the different digital content items comprises:
determining a threshold to identify and distinguish among the plurality of classifiers; and
assigning one classifier of the plurality of classifiers to each digital content item within the listing of the different digital content items based on the threshold.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein utilizing the first classification algorithm comprises utilizing a Naive Bayes algorithm.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein utilizing the second classification algorithm comprises generating a statistical probability of the demographic label for the user based on:
a number of sessions for the user;
the user-level-latent feature for the user;
the session-level features;
the first coefficient learned by the supervised learning model; and
the second coefficient learned by the supervised learning model.

### Claim 8 (depends on 7)

The computer-implemented method of claim 7, wherein utilizing the second classification algorithm further comprises training the supervised learning model based on:
a user-level-latent feature for each verified user of a plurality of verified users corresponding to verified demographic labels;
session-level features in which a verified user of the plurality of verified users consumed one or more training digital content items; and
additional demographic-label statistics comprising the verified demographic label for each verified user of the plurality of verified users.

### Claim 9 (depends on 7)

The computer-implemented method of claim 7, wherein the second classification algorithm comprises a logistic regression and the supervised learning model comprises a logistic regression model.

### Claim 10

A non-transitory computer readable storage medium storing instructions thereon that, when executed by at least one processor, cause a computing device to:
receive demographic-label statistics associated with digital content items;
access session-level data associated with the digital content items, wherein the session-level data comprises session-level features for sessions in which a user of a plurality of users consumed one or more of the digital content items;
utilize a first classification algorithm to determine a user-level-latent feature for the user by analyzing subsets of the demographic-label statistics particular to different digital content items consumed by the user and a listing of the different digital content items consumed by the user;
generate a first coefficient learned by a supervised learning model for user-level-latent features processed by the supervised learning model;
generate a second coefficient learned by the supervised learning model for session-level features processed by the supervised learning model; and
utilize a second classification algorithm to determine a demographic label for the user by analyzing the user-level-latent feature for the user, the session-level features, the first coefficient learned by the supervised learning model, and the second coefficient learned by the supervised learning model.

### Claim 11 (depends on 10)

The non-transitory computer readable storage medium of claim 10, wherein the plurality of users comprises a subset of users from a larger group of users.

### Claim 12 (depends on 10)

The non-transitory computer readable storage medium of claim 10, further comprising instructions that, when executed by the at least one processor, cause the computing device to generate a recommendation based on the demographic label for the user.

### Claim 13 (depends on 12)

The non-transitory computer readable storage medium of claim 12, wherein the recommendation comprises a suggested digital content item for the user, the suggested digital content item comprising a suggested advertisement, an application, an audio content item, a video content item, or a webpage.

### Claim 14 (depends on 10)

The non-transitory computer readable storage medium of claim 10, further comprising instructions that, when executed by the at least one processor, cause the computing device to generate a demographic-label report indicating the demographic label for the user.

### Claim 15 (depends on 10)

The non-transitory computer readable storage medium of claim 10, further comprising instructions that, when executed by the at least one processor, cause the computing device to access the session-level data by:
extracting the session-level data from an analytics log; and
aggregating the session-level data within a database.

### Claim 16

A system for determining demographic labels using digital inputs within a predictive model for demographic classification comprising:
memory comprising demographic-label statistics for digital content items; and
a server computing device comprising instructions thereon that, when executed by the server computing device, cause the system to:
access the demographic-label statistics for the digital content items;
access session-level data concerning the digital content items, wherein the session-level data comprises session-level features for sessions in which a user of a plurality of users consumed one or more of the digital content items;
utilize a first classification algorithm to determine a prior probability of a possible demographic label for the user by identifying, from within the demographic-label statistics and for each digital content item consumed by the user, a collective probability of the possible demographic label for consumers of each digital content item consumed by the user and determining a product of each collective probability of the possible demographic label;
generate a first coefficient learned by a supervised learning model for prior probabilities of possible demographic labels processed by the supervised learning model;
generate a second coefficient learned by the supervised learning model for session-level features processed by the supervised learning model; and
utilize a second classification algorithm to determine a statistical probability of a demographic label for the user by analyzing a number of sessions for the user, the prior probability of the possible demographic label for the user, the session-level features, the first coefficient learned by the supervised learning model, and the second coefficient learned by the supervised learning model.

### Claim 17 (depends on 16)

The system of claim 16, further comprising instructions that, when executed by the server computing device, cause the system to receive updated demographic-label statistics for the digital content items.

### Claim 18 (depends on 17)

The system of claim 17, further comprising instructions that, when executed by the server computing device, cause the system to utilize the first classification algorithm to re-determine the prior probability of the possible demographic label for the user based on the updated demographic-label statistics.

### Claim 19 (depends on 18)

The system of claim 18, further comprising instructions that, when executed by the server computing device, cause the system to utilize the second classification algorithm to re-determine the statistical probability of the demographic label for the user by using as an input for the second classification algorithm the re-determined prior probability of the possible demographic label for the user.

### Claim 20 (depends on 16)

The system of claim 16, wherein the demographic label comprises a first user characteristic and a second user characteristic.