# US 2024/0202550 A1 · Machine learning modeling using social graph signals

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

- Number: US 2024/0202550 A1
- Document ID: US20240202550A1
- Status: Published application
- Kind: A1
- Application number: 18/592,154
- Filed: 2024-02-29
- Published: 2024-06-20
- Earliest claimed: 2017-07-21
- Assignees: Snap Inc.
- Applicants: Snap Inc.
- Inventors: John Cain Blackwood; Jason Brewer; Nima Khajehnouri; Hadi Minooei; Benjamin C. Steele; Qian You
- Web page: https://patentgrep.com/patent/US20240202550A1

## CPC

- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06F 16/951: Indexing; Web crawling techniques
- G06F 16/9536: Search customisation based on social or collaborative filtering
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and methods are provided for receiving a request for lookalike data, the request for lookalike data comprising seed data and generating sample data from the seed data and from user data for a plurality of users, to use in a lookalike model training. The systems and methods further provide for capturing a snapshot of social graph data for a plurality of users and computing social graph features based on the seed data and the user data for the plurality of users, training a lookalike model based on the sample data and the computed social graph features to generate a trained lookalike model, generating a lookalike score for each user of the plurality of users in the user data using the trained lookalike model, and generating a list comprising a unique identifier for each user of the plurality of users and an associated lookalike score for each unique identifier.

## Claims

### Claim 1

A method comprising:
generating, by a server computer system, sample data from seed data and from user data for a plurality of users, to use in a lookalike model training, by performing operations comprising:
generating a positive data sample from the seed data to use in the lookalike model training; and
generating a negative data sample from user data stored in a database to use in the lookalike model training;
training, by the server computer system, a lookalike model based on the sample data and user profile features for the plurality of users to generate a trained lookalike model;
generating, by the server computer system, a lookalike score for each user of the plurality of users in the user data using the trained lookalike model;
generating, by the server computer system, a social graph score for each user of the plurality of users by comparing each user of the plurality of users to each user of the seed data based on at least one social graph feature;
performing, by the server computer system, a probability calculation based on the generated lookalike score for each user and the social graph score for each user;
generating, by the server computer system, a list of a subset of the plurality of users based on the probability calculation; and
generating, by the server computer system, a final lookalike score for each user in the list of the subset of the plurality of users based on the generated lookalike score for each user and the social graph score for each user.

### Claim 2 (depends on 1)

The method of claim 1, wherein the list of the subset of the plurality of users is generated based on the probability calculation by:
determining which users should be eliminated based on their respective probability calculation; and
for each user to be eliminated, setting an associated social graph score to zero and removing the user from the plurality of users.

### Claim 3 (depends on 1)

The method of claim 1, wherein the positive data sample comprises the seed data.

### Claim 4 (depends on 1)

The method of claim 1, wherein the negative data sample comprises a subset of the user data for the plurality of users.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
capturing a user profile snapshot;
generating user profile feature data; and
storing the user profile feature data.

### Claim 6 (depends on 1)

The method of claim 1, further comprising:
filtering the list of the subset of plurality of users comprising each user and an associated final lookalike score for each user, based on filter characteristics received in a request for lookalike data; and
associating the filtered list with the request for lookalike data.

### Claim 7 (depends on 1)

The method of claim 1, wherein the seed data comprises a plurality of user identifiers.

### Claim 8 (depends on 1)

The method of claim 1, the list of the subset of the plurality of users comprising a unique identifier for each user of the list of the subset of the plurality of users and an associated final lookalike score for each unique identifier is of a size indicated by a request for lookalike data.

### Claim 9 (depends on 1)

The method of claim 1, wherein the plurality of users are users of a messaging system or social networking system.

### Claim 10

A system comprising:
one or more hardware processors; and
a computer-readable medium coupled with the one or more hardware processors, the computer-readable medium comprising instructions stored thereon that are executable by the one or more hardware processors to cause the system to perform operations comprising:
generating sample data from seed data and from user data for a plurality of users, to use in a lookalike model training, by performing operations comprising:
generating a positive data sample from the seed data to use in the lookalike model training; and
generating a negative data sample from user data stored in a database to use in the lookalike model training;
training a lookalike model based on the sample data and user profile features for the plurality of users to generate a trained lookalike model;
generating a lookalike score for each user of the plurality of users in the user data using the trained lookalike model;
generating a social graph score for each user of the plurality of users by comparing each user of the plurality of users to each user of the seed data based on at least one social graph feature;
performing a probability calculation based on the generated lookalike score for each user and the social graph score for each user;
generating a list of a subset of the plurality of users based on the probability calculation; and
generating a final lookalike score for each user in the list of the subset of the plurality of users based on the generated lookalike score for each user and the social graph score for each user.

### Claim 11 (depends on 10)

The system of claim 10, wherein the list of the subset of the plurality of users is generated based on the probability calculation by:
determining which users should be eliminated based on their respective probability calculation; and
for each user to be eliminated, setting an associated social graph score to zero and removing the user from the plurality of users.

### Claim 12 (depends on 10)

The system of claim 10, wherein the positive data sample comprises the seed data.

### Claim 13 (depends on 10)

The system of claim 10, wherein the negative data sample comprises a subset of the user data for the plurality of users.

### Claim 14 (depends on 10)

The system of claim 10, the operations further comprising:
capturing a user profile snapshot;
generating user profile feature data; and
storing the user profile feature data.

### Claim 15 (depends on 10)

The system of claim 10, the operations further comprising:
filtering the list of the subset of plurality of users comprising each user and an associated final lookalike score for each user, based on filter characteristics received in a request for lookalike data; and
associating the filtered list with the request for lookalike data.

### Claim 16 (depends on 10)

The system of claim 10, wherein the seed data comprises a plurality of user identifiers.

### Claim 17 (depends on 10)

The system of claim 10, the list of the subset of the plurality of users comprising a unique identifier for each user of the list of the subset of the plurality of users and an associated final lookalike score for each unique identifier is of a size indicated by a request for lookalike data.

### Claim 18 (depends on 10)

The system of claim 10, wherein the plurality of users are users of a messaging system or social networking system.

### Claim 19

A non-transitory computer-readable medium comprising instructions stored thereon that are executable by at least one processor to cause a computing device to perform operations comprising:
generating sample data from seed data and from user data for a plurality of users, to use in a lookalike model training, by performing operations comprising:
generating a positive data sample from the seed data to use in the lookalike model training; and
generating a negative data sample from user data stored in a database to use in the lookalike model training;
training a lookalike model based on the sample data and user profile features for the plurality of users to generate a trained lookalike model;
generating a lookalike score for each user of the plurality of users in the user data using the trained lookalike model;
generating a social graph score for each user of the plurality of users by comparing each user of the plurality of users to each user of the seed data based on at least one social graph feature;
performing a probability calculation based on the generated lookalike score for each user and the social graph score for each user;
generating a list of a subset of the plurality of users based on the probability calculation; and
generating a final lookalike score for each user in the list of the subset of the plurality of users based on the generated lookalike score for each user and the social graph score for each user.

### Claim 20 (depends on 19)

The non-transitory computer-readable medium of claim 19, wherein the list of the subset of the plurality of users is generated based on the probability calculation by:
determining which users should be eliminated based on their respective probability calculation; and
for each user to be eliminated, setting an associated social graph score to zero and removing the user from the plurality of users.
