# US 12,602,719 B2 · Seasonality score systems and methods

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

- Number: US 12,602,719 B2
- Document ID: US12602719B2
- Status: Granted
- Kind: B2
- Application number: 18/791,842
- Filed: 2024-08-01
- Published: 2024-11-28
- Granted: 2026-04-14
- Earliest claimed: 2021-12-20
- Assignees: Walmart Apollo, LLC
- Applicants: Walmart Apollo, LLC
- Inventors: Luyi Ma; Nimesh Sinha; Parth Ramesh Vajge; Hyun Duk Cho; Sushant Kumar; Kannan Achan
- Web page: https://patentgrep.com/patent/US12602719B2

## CPC

- G06Q 30/0631: {Recommending goods or services}
- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06Q 30/0202: Market predictions or forecasting for commercial activities
- G06Q 30/0204: Market segmentation
- G06Q 30/0256: {User search}
- G06Q 30/0641: {utilising user interfaces specially adapted for shopping}
- 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

Systems and methods for generating and using seasonal affinity scores is disclosed. A set of user-specific historical transaction data is obtained and a user-specific affinity score including at least one of a user-specific season affinity score or a user-specific seasonal theme affinity score is determined by determining one or more product affinity scores for a set of product taxonomies and combining the one or more product affinity scores with one or more product index scores to generate the user-specific affinity score. The product affinity scores are determined by a trained scoring calculation model configured to receive the set of user-specific historical transaction data. One or more interface elements are selected based on the user-specific affinity score and an interface is generated including the one or more interface elements.

## Claims

### Claim 1

A system, comprising:
a processor; and
a non-transitory memory storing instructions, that when executed, cause the processor to:
receive, from a database, a set of user-specific historical transaction data;
receive, from the database, a training data set including a set of historical transaction data associated with a plurality of users over a predetermined time period;
iteratively train a first machine learning model based on implicit feedback data and the training data set to find season affinity scores by minimizing a first cost function that includes a first parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the first cost function comprises generating each of a first user matrix and a first product type matrix based on the training data, inputting the first user matrix and the first product type matrix into the first cost function, and comparing, by the first cost function, the first parameter with the first user matrix and the first product type matrix;
iteratively train a second machine learning model based on implicit feedback data and the training data set to find product affinity scores by minimizing a second cost function that includes a second parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the second cost function comprises generating each of a second user matrix and a second product type matrix based on the training data, inputting the second user matrix and the second product type matrix into the second cost function, and comparing, by the second cost function, the second parameter with the second user matrix and the second product type matrix;
input at least first portions of the user-specific historical transaction data to the trained first machine learning model and, in response, generate user-specific seasonal product affinity scores;
generate a user-specific season affinity score based on the user-specific seasonal product affinity scores;
input at least second portions of the user-specific historical transaction data to the trained second machine learning model and, in response, generate user-specific product affinity scores;
generate a plurality of user-specific seasonal theme affinity scores based on the user-specific product affinity scores and seasonal product index scores;
determine a user-specific affinity score associated with the specific user based on the user-specific season affinity score and the plurality of user-specific seasonal theme affinity scores;
compare the user-specific affinity score to a predetermined dynamic threshold, the predetermined dynamic threshold being based on one or more external factors associated with the specific user; and
generate, on a user interface, one or more user-specific interface elements if the user-specific affinity score is above the predetermined dynamic threshold and one or more generic interface elements if the user-specific affinity score is below the predetermined dynamic threshold.

### Claim 2 (depends on 1)

The system of claim 1, wherein the set of user-specific historical transaction data is filtered based on a set of seasonal products taxonomies prior to being provided to the trained first machine learning model and the trained second machine learning model.

### Claim 3 (depends on 1)

The system of claim 1, wherein:
each of the first machine learning model and the second machine learning model is a factorization model configured to alternate between solving for a first matrix by fixing a second matrix and solving for the second matrix by fixing the first matrix.

### Claim 4 (depends on 1)

The system of claim 1, wherein:
the second cost function comprises a preference for a user to an item determined based on a number of transactions, a set of variables measuring confidence in observing the preference, a column in a user matrix, and a column in a product type matrix.

### Claim 5 (depends on 1)

The system of claim 1, wherein each of the first machine learning model and the second machine learning model is a trained alternating least square (ALS) model.

### Claim 6 (depends on 1)

The system of claim 1, wherein the user-specific affinity score is determined by:
obtaining a set of season to product mappings identifying product taxonomies associated with one or more selected seasons;
inputting the user-specific historical transaction data into the trained first machine learning model to generate user-specific seasonal product affinity scores;
combining the user-specific seasonal product affinity scores with seasonal product index scores to generate the user-specific season affinity score;
obtaining a set of seasonal theme to product type mappings identifying product taxonomies associated with one or more seasonal themes of the one or more selected seasons;
inputting the user-specific historical transaction data into the trained second machine learning model to generate user-specific product affinity scores;
combining the user-specific product affinity scores with seasonal product index scores to generate the plurality of user-specific seasonal theme affinity scores; and
computing the user-specific affinity score associated with the specific user based on the user-specific season affinity score and the plurality of user-specific seasonal theme affinity scores.

### Claim 7 (depends on 6)

The system of claim 6, wherein combining the user-specific seasonal product affinity scores with the seasonal product index scores comprises a weighted combination or a dot combination.

### Claim 8 (depends on 1)

The system of claim 1, wherein the one or more user-specific interface elements are generated based on:
selecting interface elements associated with a season when the user-specific affinity score is above a predetermined threshold; and
selecting generic interface elements when the user-specific affinity score is below the predetermined threshold.

### Claim 9

A computer-implemented method, comprising:
receiving, from a database, a set of user-specific historical transaction data;
receiving, from the database, a training data set including a set of historical transaction data associated with a plurality of users over a predetermined time period;
iteratively training a first machine learning model based on implicit feedback data and the training data set to find season affinity scores by minimizing a first cost function that includes a first parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the first cost function comprises generating each of a first user matrix and a first product type matrix based on the training data, inputting the first user matrix and the first product type matrix into the first cost function, and comparing, by the first cost function, the first parameter with the first user matrix and the first product type matrix;
iteratively training a second machine learning model based on implicit feedback data and the training data set to find product affinity scores by minimizing a second cost function that includes a second parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the second cost function comprises generating each of a second user matrix and a second product type matrix based on the training data, inputting the second user matrix and the second product type matrix into the second cost function, and comparing, by the second cost function, the second parameter with the second user matrix and the second product type matrix;
inputting at least first portions of the user-specific historical transaction data to the trained first machine learning model and, in response, generating user-specific seasonal product affinity scores;
generating a user-specific season affinity score based on the user-specific seasonal product affinity scores;
inputting at least second portions of the user-specific historical transaction data to the trained second machine learning model and, in response, generating user-specific product affinity scores;
generating a plurality of user-specific seasonal theme affinity scores based on the user-specific product affinity scores and seasonal product index scores;
determining a user-specific affinity score associated with the specific user based on the user-specific season affinity score and the plurality of user-specific seasonal theme affinity scores;
comparing the user-specific affinity score to a predetermined dynamic threshold, the predetermined dynamic threshold being based on one or more external factors associated with the specific user; and
generating, on a user interface, one or more user-specific interface elements if the user-specific affinity score is above the predetermined dynamic threshold and one or more generic interface elements if the user-specific affinity score is below the predetermined dynamic threshold.

### Claim 10 (depends on 9)

The method of claim 9, wherein the set of user-specific historical transaction data is filtered based on a set of seasonal products taxonomies prior to being provided to the trained first machine learning model and the trained second machine learning model.

### Claim 11 (depends on 9)

The method of claim 9, wherein:
each of the first machine learning model and the second machine learning model is a factorization model configured to alternate between solving for a first matrix by fixing a second matrix and solving for the second matrix by fixing the first matrix.

### Claim 12 (depends on 9)

The method of claim 9, wherein:
the second cost function comprises a preference for a user to an item determined based on a number of transactions, a set of variables measuring confidence in observing the preference, a column in a user matrix, and a column in a product type matrix.

### Claim 13 (depends on 9)

The method of claim 9, wherein each of the first machine learning model and the second machine learning model is a trained alternating least square (ALS) model.

### Claim 14 (depends on 9)

The method of claim 9, wherein determining the user-specific affinity score comprises:
obtaining a set of season to product mappings identifying product taxonomies associated with one or more selected seasons;
inputting the user-specific historical transaction data into the trained first machine learning model to generate user-specific seasonal product affinity scores;
combining the user-specific seasonal product affinity scores with seasonal product index scores to generate the user-specific season affinity score;
obtaining a set of seasonal theme to product type mappings identifying product taxonomies associated with one or more seasonal themes of the one or more selected seasons;
inputting the user-specific historical transaction data into the trained second machine learning model to generate user-specific product affinity scores;
combining the user-specific product affinity scores with seasonal product index scores to generate the plurality of user-specific seasonal theme affinity scores; and
computing the user-specific affinity score associated with the specific user based on the user-specific season affinity score and the plurality of user-specific seasonal theme affinity scores.

### Claim 15 (depends on 14)

The method of claim 14, wherein combining the user-specific seasonal product affinity scores with the seasonal product index scores comprises a weighted combination or a dot combination.

### Claim 16 (depends on 9)

The method of claim 9, wherein the one or more user-specific interface elements are generated based on:
selecting interface elements associated with a season when the user-specific affinity score is above a predetermined threshold; and
selecting generic interface elements when the user-specific affinity score is below the predetermined threshold.

### Claim 17

A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by a processor cause a device to perform operations comprising:
receiving, from a database, a set of user-specific historical transaction data;
receiving, from the database, a training data set including a set of historical transaction data associated with a plurality of users over a predetermined time period;
iteratively training a first machine learning model based on implicit feedback data and the training data set to find season affinity scores by minimizing a first cost function that includes a first parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the first cost function comprises generating each of a first user matrix and a first product type matrix based on the training data, inputting the first user matrix and the first product type matrix into the first cost function, and comparing, by the first cost function, the first parameter with the first user matrix and the first product type matrix;
iteratively training a second machine learning model based on implicit feedback data and the training data set to find product affinity scores by minimizing a second cost function that includes a second parameter characterizing preferences of users to items based on a number of corresponding transactions, and wherein minimizing the second cost function comprises generating each of a second user matrix and a second product type matrix based on the training data, inputting the second user matrix and the second product type matrix into the second cost function, and comparing, by the second cost function, the second parameter with the second user matrix and the second product type matrix;
inputting at least first portions of the user-specific historical transaction data to the trained first machine learning model and, in response, generating user-specific seasonal product affinity scores;
generating a user-specific season affinity score based on the user-specific seasonal product affinity scores;
inputting at least second portions of the user-specific historical transaction data to the trained second machine learning model and, in response, generating user-specific product affinity scores;
generating a plurality of user-specific seasonal theme affinity scores based on the user-specific product affinity scores and seasonal product index scores;
determining a user-specific affinity score associated with the specific user based on the user-specific season affinity score and the plurality of user-specific seasonal theme affinity scores;
comparing the user-specific affinity score to a predetermined dynamic threshold, the predetermined dynamic threshold being based on one or more external factors associated with the specific user; and
generating, on a user interface, one or more user-specific interface elements if the user-specific affinity score is above the predetermined dynamic threshold and one or more generic interface elements if the user-specific affinity score is below the predetermined dynamic threshold.

### Claim 18 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein the set of user-specific historical transaction data is filtered based on a set of seasonal products taxonomies prior to being provided to the trained first machine learning model and the trained second machine learning model.

### Claim 19 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein:
each of the first machine learning model and the second machine learning model is a factorization model configured to alternate between solving for a first matrix by fixing a second matrix and solving for the second matrix by fixing the first matrix.

### Claim 20 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein:
the second cost function comprises a preference for a user to an item determined based on a number of transactions, a set of variables measuring confidence in observing the preference, a column in a user matrix, and a column in a product type matrix.