Cross-system recommender

Patent metadata

Field	Value
Publication number	US-12493906-B2
Application number	US-202418593624-A
Country	US
Kind code	B2
Filing date	Mar 1, 2024
Priority date	Mar 1, 2024
Publication date	Dec 9, 2025
Grant date	Dec 9, 2025

Abstract

A method for generating cross-channel recommendations for a customer includes receiving customer data associated with the customer, content data, and clickstream data; encoding the content data using a text encoder, the encoding resulting in content embeddings; encoding the clickstream data using a clickstream encoder, the encoding resulting in clickstream embeddings; providing the content data, the clickstream data, and the clickstream embeddings as inputs to a hybrid latent model; causing execution of the hybrid latent model, the execution resulting in cross-system user and item interaction embeddings; retrieving application features corresponding to the clickstream data from an application feature store; providing the customer data, the content embeddings, the clickstream embeddings, the cross-system user and item interaction embeddings, and the application features as inputs to a recommendation model; causing execution of the recommendation model, the execution resulting in a plurality of ranked recommendations.

First claim

What is claimed is:

1 . A computer-implemented method for generating a plurality of cross-channel recommendations for a customer, the method comprising:
receiving, by a computer system, customer data associated with the customer, content data, and clickstream data; encoding, by the computer system, the content data using a text encoder, the encoding resulting in content embeddings; encoding, by the computer system, the clickstream data using a clickstream encoder, the encoding resulting in clickstream embeddings; providing, by the computer system, one or more of the content data, the clickstream data, user-item interaction data, the customer data, and clickstream embeddings as inputs to a hybrid latent model; causing, by the computer system, execution of the hybrid latent model, the execution resulting in cross-channel interaction embeddings; retrieving, by the computer system, application features corresponding to the clickstream data from an application feature store; providing, by the computer system, the customer data, the clickstream embeddings, the content embeddings, the cross-system interaction embeddings, the user-item interaction data, and the application features as inputs to a recommendation model; and causing, by the computer system, execution of the recommendation model, the execution resulting in a plurality of ranked recommendations.

2 . The computer-implemented method of claim 1 , wherein the clickstream data includes data from a plurality of channels and at least one channel of the plurality of channels has a different type from another channel of the plurality of channels.

3 . The computer-implemented method of claim 1 , wherein the clickstream data includes at least one of clickstream data associated with the customer and a channel and clickstream data associated with the customer within a timeframe.

4 . The computer-implemented method of claim 1 , wherein the customer data includes at least one of demographic data, account data, and transaction data.

5 . The computer-implemented method of claim 1 , wherein the application features include at least one of click-through-rate, number of unique customers, number of unique presented items, and usage volume.

6 . The computer-implemented method of claim 1 , further comprising filtering, by the computer system, the plurality of ranked recommendations.

7 . The computer-implemented method of claim 1 , further comprising re-ranking, by the computer system, the plurality of ranked recommendations.

8 . The computer-implemented method of claim 1 , wherein the hybrid latent model includes a matrix factorization framework.

9 . The computer-implemented method of claim 1 , wherein the recommendation model uses gradient boosting.

10 . The computer-implemented method of claim 1 , further comprising causing, by the computer system, display of at least one of the plurality of recommendations to the customer.

11 . A system for generating a plurality of cross-channel recommendations for a customer, the system comprising:
a hybrid latent model; a recommendation model; and a computer system having a processor coupled to a memory, the computer system communicatively coupled to the hybrid latent model and to the recommendation model, the processor configured to:
receive customer data associated with the customer, content data, and clickstream data;
encode the content data using a text encoder, the encoding resulting in content embeddings;
encode the clickstream data using a clickstream encoder, the encoding resulting in clickstream embeddings;
provide the customer data, the clickstream data, and the clickstream embeddings as inputs to the hybrid latent model;
cause execution of the hybrid latent model, the execution resulting in cross-system interaction embeddings;
retrieve application features corresponding to the clickstream data from an application feature store;
provide the customer data, the clickstream embeddings, the content embeddings, the cross-system interaction embeddings, user-item interaction data, and the application features as inputs to a recommendation model; and
cause execution of the recommendation model, the execution resulting in a plurality of ranked recommendations.

12 . The system of claim 11 , wherein the clickstream data includes data from a plurality of channels and at least one channel of the plurality of channels has a different type from another channel of the plurality of channels.

13 . The system of claim 11 , wherein the clickstream data includes at least one of clickstream data associated with the customer and a web page and clickstream data associated with the customer within a timeframe.

14 . The system of claim 11 , wherein the customer data includes at least one of demographic data, account data, and transaction data.

15 . The system of claim 11 , wherein the processor is further configured to filter the plurality of ranked recommendations.

16 . The system of claim 11 , wherein the processor is further configured to re-rank the plurality of ranked recommendations.

17 . The system of claim 11 , wherein the application features include at least one of click-through-rate, number of unique customers, number of unique presented items, and usage volume.

18 . The system of claim 11 , wherein the hybrid latent model includes a matrix factorization framework.

19 . The system of claim 11 , wherein the recommendation model uses gradient boosting.

20 . A non-transitory computer-readable medium having software encoded thereon, the software, when executed by a computer system, operable to:
receive customer data associated with a customer, content data, and clickstream data; encode the content data using a text encoder, the encoding resulting in content embeddings; encode the clickstream data using a clickstream encoder, the encoding resulting in clickstream embeddings; provide the content data, the clickstream data, and one or more of the clickstream embeddings and customer data as inputs to a hybrid latent model; cause execution of the hybrid latent model, the execution resulting in cross-system interaction embeddings; retrieve application features corresponding to the clickstream data from an application feature store; provide the customer data, the content embeddings, the clickstream embeddings, the cross-system interaction embeddings, and the application features as inputs to a recommendation model; and cause execution of the recommendation model, the execution resulting in a plurality of ranked recommendations.

Assignees

• Fmr Llc

Inventors

• Kleynhans Bernard
• Kadioglu Serdar
• Baral Ramesh

Classifications

• G06Q30/0204

  Market segmentation · CPC title

• G06Q30/0631 Primary

  Recommending goods or services · CPC title