# US 2026/0073292 A1 · Systems And Methods for Automatic Treatment Recommendation For Digital Platform Display Using Machine Learning Techniques

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

- Number: US 2026/0073292 A1
- Document ID: US20260073292A1
- Status: Published application
- Kind: A1
- Application number: 19/070,387
- Filed: 2025-03-04
- Published: 2026-03-12
- Earliest claimed: 2024-09-06
- Assignees: Auxia, Inc.
- Applicants: Auxia, Inc.
- Inventors: Ravi Desu; Sumeet Kumar; Asim Krishna Prasad; Yichen Zhao
- Web page: https://patentgrep.com/patent/US20260073292A1

## CPC

- G06N 20/00: Machine learning
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06Q 30/0255: {based on user history}
- G06Q 30/0269: {based on user profile or attribute}
- G06Q 30/0271: {Personalized advertisement}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- 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

## Abstract

Disclosed herein is a computerized method including operations of obtaining user attributes and user event data, generating user embeddings and aggregated user features from the user event data and the user attributes, obtaining treatment attributes and a set of treatments corresponding to the treatment attributes, generating treatment embeddings and aggregated treatment features from the set of treatments and the treatment attributes, and generating a trained machine learning model by processing the user embeddings, the aggregated user features, the treatment embeddings, and the aggregated treatment features by a machine learning algorithm, wherein the trained machine learning machine is configured to generate a score for each treatment of the set of treatments indicative of a likelihood that serving of a particular treatment will result in performance of an objective. The user event data may include event sequence data indicating a sequence of user input actions corresponding to one or more treatments.

## Claims

### Claim 1

A computerized method comprising:
obtaining user attributes and user event data;
generating user embeddings and aggregated user features from the user event data and the user attributes;
obtaining treatment attributes and a set of treatments corresponding to the treatment attributes;
generating treatment embeddings and aggregated treatment features from the set of treatments and the treatment attributes; and
generating a trained machine learning model by processing the user embeddings, the aggregated user features, the treatment embeddings, and the aggregated treatment features by a machine learning algorithm, wherein the trained machine learning machine is configured to generate a score for each treatment of the set of treatments indicative of a likelihood that serving of a particular treatment will result in performance of an objective.

### Claim 2 (depends on 1)

The computerized method of claim 1, wherein the user event data includes event sequence data indicating a sequence of user input actions corresponding to one or more treatments.

### Claim 3 (depends on 1)

The computerized method of claim 1, wherein generating the user embeddings is performed by an autoencoder architecture, wherein the user embeddings corresponds to a latent representation generated by an encoder of the autoencoder architecture.

### Claim 4 (depends on 1)

The computerized method of claim 1, wherein generating the treatment embeddings includes operations of parsing the set of treatments and extracting features therefrom.

### Claim 5 (depends on 1)

The computerized method of claim 1, wherein generating the treatment embeddings is performed by an autoencoder architecture, wherein the treatment embeddings corresponds to a latent representation generated by an encoder of the autoencoder architecture.

### Claim 6 (depends on 1)

The computerized method of claim 1, wherein the objective is a predefined action performed through user input.

### Claim 7 (depends on 1)

The computerized method of claim 1 further comprising:
deploying the trained machine learning model including providing a set of user data, a set of candidate treatments, and the objective as input, wherein the trained machine learning model generates a set of scores for each of the set of candidate treatments; and
providing at least a first treatment of the set of candidate treatments based on a first score generated for the first treatment by the trained machine learning model.

### Claim 8

A computing device, comprising:
a processor; and
a non-transitory computer-readable medium having stored thereon instructions that, when executed by the processor, cause the processor to perform operations including:
obtaining user attributes and user event data,
generating user embeddings and aggregated user features from the user event data and the user attributes,
obtaining treatment attributes and a set of treatments corresponding to the treatment attributes,
generating treatment embeddings and aggregated treatment features from the set of treatments and the treatment attributes, and
generating a trained machine learning model by processing the user embeddings, the aggregated user features, the treatment embeddings, and the aggregated treatment features by a machine learning algorithm, wherein the trained machine learning machine is configured to generate a score for each treatment of the set of treatments indicative of a likelihood that serving of a particular treatment will result in performance of an objective.

### Claim 9 (depends on 8)

The computing device of claim 8, wherein the user event data includes event sequence data indicating a sequence of user input actions corresponding to one or more treatments.

### Claim 10 (depends on 8)

The computing device of claim 8, wherein generating the user embeddings is performed by an autoencoder architecture, wherein the user embeddings corresponds to a latent representation generated by an encoder of the autoencoder architecture.

### Claim 11 (depends on 8)

The computing device of claim 8, wherein generating the treatment embeddings includes operations of parsing the set of treatments and extracting features therefrom.

### Claim 12 (depends on 8)

The computing device of claim 8, wherein generating the treatment embeddings is performed by an autoencoder architecture, wherein the treatment embeddings corresponds to a latent representation generated by an encoder of the autoencoder architecture.

### Claim 13 (depends on 8)

The computing device of claim 8, wherein the objective is a predefined action performed through user input.

### Claim 14 (depends on 8)

The computing device of claim 8, wherein the operations further include:
deploying the trained machine learning model including providing a set of user data, a set of candidate treatments, and the objective as input, wherein the trained machine learning model generates a set of scores for each of the set of candidate treatments; and
providing at least a first treatment of the set of candidate treatments based on a first score generated for the first treatment by the trained machine learning model.

### Claim 15

A non-transitory computer-readable medium having stored thereon instructions that, when executed by one or more processors, cause the one or more processor to perform operations including:
obtaining user attributes and user event data;
generating user embeddings and aggregated user features from the user event data and the user attributes;
obtaining treatment attributes and a set of treatments corresponding to the treatment attributes;
generating treatment embeddings and aggregated treatment features from the set of treatments and the treatment attributes;
generating a trained machine learning model by processing the user embeddings, the aggregated user features, the treatment embeddings, and the aggregated treatment features by a machine learning algorithm, wherein the trained machine learning machine is configured to generate a score for each treatment of the set of treatments indicative of a likelihood that serving of a particular treatment will result in performance of an objective.

### Claim 16 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein the user event data includes event sequence data indicating a sequence of user input actions corresponding to one or more treatments.

### Claim 17 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein generating the user embeddings is performed by an autoencoder architecture, wherein the user embeddings corresponds to a latent representation generated by an encoder of the autoencoder architecture; and
wherein generating the treatment embeddings is performed by a second autoencoder architecture, wherein the treatment embeddings corresponds to a latent representation generated by an encoder of the second autoencoder architecture.

### Claim 18 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein generating the treatment embeddings includes operations of parsing the set of treatments and extracting features therefrom.

### Claim 19 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein the objective is a predefined action performed through user input.

### Claim 20 (depends on 15)

The non-transitory computer-readable medium of claim 15, wherein the operations further include:
deploying the trained machine learning model including providing a set of user data, a set of candidate treatments, and the objective as input, wherein the trained machine learning model generates a set of scores for each of the set of candidate treatments; and
providing at least a first treatment of the set of candidate treatments based on a first score generated for the first treatment by the trained machine learning model.
