System and method for obtaining recommendations using scalable cross-domain collaborative filtering

Patent metadata

Field	Value
Publication number	US-11157987-B2
Application number	US-201816215808-A
Country	US
Kind code	B2
Filing date	Dec 11, 2018
Priority date	Dec 7, 2018
Publication date	Oct 26, 2021
Grant date	Oct 26, 2021

Abstract

Aspects of the present disclosure involve systems, methods, devices, and the like for presenting a recommendation. In one embodiment, a system is introduced that includes a plurality of models for obtaining a recommendation score. The recommendation score may be obtained using one or more recommendation models and a recommendation made based on the recommendation score determined. In another embodiment, the system is introduced that can re-train the recommendation model based on a feedback received in response to a recommendation made using on the recommendation score obtained.

First claim

What is claimed is:

1. A system, comprising:
a non-transitory memory storing instructions;
a processor configured to execute the instructions to cause the system to:
in response to a notification received that a user is at a checkout, retrieve, from a customer engagement platform, user characterization information;
select one or more recommendation models to use for determining a recommendation to present to the user, based on the user characterization information,
the selected one or more recommendation models including one or both of a random walk model or a clustering model; and
the clustering model selected when the user characterization information includes user profile information;

compute, using the selected one or more recommendation models, at least one recommendation score;
analyze the at least one recommendation score to determine the recommendation to present to the user;
transmit, over a communication network and to a computing device associated with the user, the determined recommendation that causes the determined recommendation to be displayed on the computing device to the user;
receive, via the computing device and in response to the determined recommendation displayed to the user, feedback from the user; and
re-train the selected one or more recommendation models based on the received feedback in a recommendation model-training feedback loop including the selected one or more recommendation models and the received feedback.

2. The system of claim 1 , wherein:
the one or more recommendation models include a plurality of recommendation models; and
the at least one recommendation score is computed using the plurality of recommendation models and an ensemble model.

3. The system of claim 2 , wherein the ensemble model includes a decision tree model.

4. The system of claim 2 , wherein the ensemble model uses a plurality of recommendation scores generated by the plurality of recommendation models to compute the at least one recommendation score.

5. The system of claim 1 , wherein the random walk model is selected when the user characterization information includes a combination of user profile information and peer-to-peer transactions information.

6. The system of claim 1 , wherein the selected one or more recommendation models comprises at least two recommendation models that further include a cross-domain filtering model.

7. The system of claim 1 , wherein the cross-domain filtering model is selected when the user characterization information includes a combination of user profile information, peer-to-peer transactions information, and cross-domain transactional information.

8. A method performed by one or more processors, comprising:
in response to a notification received that a user is at a checkout, retrieving, from a customer engagement platform, user characterization information;
selecting one or more recommendation models to use for determining a recommendation to present to the user, based on the user characterization information,
the selected one or more recommendation models including one or both of a random walk model or a clustering model;
the random walk model is selected when the user characterization information includes a combination of user profile information and peer-to-peer transactions information; and
the clustering model selected when the user characterization information includes user profile information;

computing, using the selected one or more recommendation models, at least one recommendation score;
analyzing the at least one recommendation score to determine the recommendation to present to the user;
transmitting, over a communication network and to a computing device associated with the user, the determined recommendation that causes the determined recommendation to be presented the user through a display of the computing device;
receiving, via the computing device and in response to the determined recommendation presented to the user, feedback from the user; and
re-training the selected one or more recommendation models based on the received feedback in a recommendation model-training feedback loop including the selected one or more recommendation models and the received feedback.

9. The method of claim 8 , wherein:
the one or more recommendation models include a plurality of recommendation models; and
the at least one recommendation score is computed using the plurality of recommendation models and an ensemble model.

10. The method of claim 9 , wherein the ensemble model includes a decision tree model.

11. The method of claim 9 , wherein the ensemble model uses a plurality of recommendation scores generated by the plurality of recommendation models to compute the at least one recommendation score.

12. The method of claim 8 , wherein the selected one or more recommendation models comprises at least two recommendation models that further include a cross-domain filtering model.

13. The method of claim 8 , wherein the cross-domain filtering model is selected when the user characterization information includes a combination of user profile information, peer-to-peer transactions information, and cross-domain transactional information.

14. A non-transitory machine-readable medium having instructions stored thereon, the instructions executable to cause performance of operations comprising:
in response to a notification received that a user is at a checkout, retrieving, from a customer engagement platform, user characterization information;
selecting one or more recommendation models to use for determining a recommendation to present to the user, based on the user characterization information,
the selected one or more recommendation models including one or both of a random walk model or a clustering model; and
the clustering model selected when the user characterization information includes user profile information;

computing, using the selected one or more recommendation models, at least one recommendation score;
analyzing the at least one recommendation score to determine the recommendation to present to the user;
transmitting, over a communication network and to a computing device associated with the user, the determined recommendation to cause the determined recommendation to be presented to the user via the computing device;
receiving, via the computing device and in response to the determined recommendation presented to the user, feedback from the user; and
re-training the selected one or more recommendation models based on the received feedback in a recommendation model-training feedback loop including the selected one or more recommendation models and the received feedback.

15. The non-transitory machine-readable medium of claim 14 , wherein the one or more recommendation models include a plurality of recommendation models; and
the at least one recommendation score is computed using the plurality of recommendation models and an ensemble model.

16. The non-transitory machine-readable medium of claim 15 , wherein the ensemble model includes a decision tree model and uses a plurality of recommendation scores generated by the plurality of recommendation models to compute the at least one recommendation score.

17. The non-transitory machine-readable medium of claim 14 , wherein the random walk model is selected when the user characterization information includes a combination of user profile information and peer-to-peer transactions information.

18. The non-transitory machine-readable medium of claim 14 , wherein the selected one or more recommendation models comprises

Assignees

• Paypal Inc

Inventors

• Kumar Dinesh
• Pan Yuanyuan
• Gaurav Prashant
• Kurniadi Fransisco
• Xin Yue
• Govindarajalu Krishnakumar
• Kidney Kimberly
• Sun Tao
• Ward Kevin

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N5/048

  Fuzzy inferencing · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N5/04

  Inference or reasoning models · CPC title