IP Library Granted Patent US 12700003
Granted Patent B2
US 12700003 · App. 17/804,537 · Granted Aug 4, 2026

Systems and methods for frequent machine learning model retraining and rule optimization

Inventors: Mike Hughes (Delaware, DE); Yea Kang Yoon (Wilmington, DE); Fang-Yu Lin (New York, NY); Ramana Nallajarla (New York, NY); Sambasiva R Vadlamudi (Garnet Valley, NY); Josh X Jiang (New York, NY); Hari Sivaprasad (New York, NY); Benedict Hall (New York, NY); Lifeng Wang (New York, NY)
Assignee: JPMORGAN CHASE BANK, N.A.
G06Q20/4016G06N20/00G06N20/20
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12700003
App. No.
17/804,537
Granted
Aug 4, 2026
Kind
B2
Abstract

Systems and methods for frequent machine learning model retraining and rule optimization are disclosed. In accordance with aspects, a method may include generating a challenger machine learning model based on a production machine learning model; training the challenger machine learning model on a plurality of datasets; scoring historical data with the challenger machine learning model, wherein the scoring produces a respective score for each record of a plurality of records in the historical data; determining that the challenger model performs within predetermined thresholds based on the scoring; selecting an optimal scaler value for a rule based on execution of the rule with a range of scaler values applied to the respective score for each record of the plurality of records evaluated by the rule; determining that the optimal scaler value outperforms a production scaler value; and promoting the challenger model and the optimal scaler value to a production environment.

Claims (41)

1 . A method for frequent machine learning model retraining and rule optimization in a payment transaction fraud detection system implemented by at least one computing device, comprising:

generating a challenger machine learning model based on a production machine learning model;

training the challenger machine learning model on a plurality of datasets, wherein each of the plurality of datasets includes data records collected on a date defined as a number of days previous to a current date, wherein the plurality of datasets include historical payment transaction data;

scoring the historical payment transaction data with the challenger machine learning model, wherein the scoring produces a respective score for each record of a plurality of records in the plurality of datasets, wherein the scoring produces respective fraud scores for said each record of the plurality of records;

determining that the challenger model performs within predetermined thresholds based on the scoring;

selecting an optimal scaler value for a rule based on execution of the rule with a range of scaler values applied to the respective score for said each record of the plurality of records evaluated by the rule;

determining that the optimal scaler value outperforms a production scaler value; and

promoting the challenger model and the optimal scaler value to a production environment to automatically process payment transaction data for fraud detection by (1) calculating a model score of the payment transaction data and assigning a return on investment value to the payment transaction data based on variables of the payment transaction data and (2) calculating a fraud score of the payment transaction data.

2 . The method of claim 1 , wherein the plurality of datasets includes a first dataset, and wherein the first dataset includes data records collected on a date defined as 1 day previous to the current day;

wherein the plurality of datasets includes a second dataset, and wherein the second dataset includes data records collected on a date defined as 14 days previous to the current day; and

wherein the plurality of datasets includes a third dataset and wherein the third dataset includes data records collected on a date defined as 90 days previous to the current day.

3 . The method of claim 1 , wherein the rule is a segment rule for a segment, and wherein the segment is defined by a transaction type.

4 . The method of claim 3 , wherein the segment is one of a plurality of segments, and where each segment of the plurality of segments is defined by a different rule type.

5 . The method of claim 4 , wherein said each segment of the plurality of segments includes a segment rule and wherein said each segment rule includes a scaler value of which can be set independently.

6 . The method of claim 1 , wherein the plurality of datasets includes data collected within a predetermined time window.

7 . The method of claim 6 , wherein the predetermined time window is from the date defined as 1 day previous to the current date to the date defined as 14 days previous to the current date.

8 . The method of claim 1 , wherein the challenger machine learning model is based on an extreme gradient boosting algorithm.

9 . A system for frequent machine learning model retraining and rule optimization in a payment transaction fraud detection system comprising at least one computing device including a processor, wherein the at least one computing device is configured to:

generate a challenger machine learning model based on a production machine learning model;

train the challenger machine learning model on a plurality of datasets, wherein each of the plurality of datasets includes data records collected on a date defined as a number of days previous to a current date, wherein the plurality of datasets include historical payment transaction data;

score the historical payment transaction data with the challenger machine learning model, wherein the scoring produces a respective score for each record of a plurality of records in the plurality of datasets, wherein the scoring produces respective fraud scores for said each record of the plurality of records;

determine that the challenger model performs within predetermined thresholds based on the scoring;

select an optimal scaler value for a rule based on execution of the rule with a range of scaler values applied to the respective score for said each record of the plurality of records evaluated by the rule;

determine that the optimal scaler value outperforms a production scaler value; and

promote the challenger model and the optimal scaler value to a production environment to automatically process payment transaction data for fraud detection by (1) calculating a model score of the payment transaction data and assigning a return on investment value to the payment transaction data based on variables of the payment transaction data and (2) calculating a fraud score of the payment transaction data.

10 . The system of claim 9 , wherein the plurality of datasets includes a first dataset, and wherein the first dataset includes data records collected on a date defined as 1 day previous to the current day;

wherein the plurality of datasets includes a second dataset, and wherein the second dataset includes data records collected on a date defined as 14 days previous to the current day; and

wherein the plurality of datasets includes a third dataset and wherein the third dataset includes data records collected on a date defined as 90 days previous to the current day.

11 . The method of claim 9 , wherein the rule is a segment rule for a segment, and wherein the segment is defined by a transaction type.

12 . The system of claim 11 , wherein the segment is one of a plurality of segments, and where each segment of the plurality of segments is defined by a different rule type.

13 . The system of claim 12 , wherein said each segment of the plurality of segments includes a segment rule and wherein said each segment rule includes a scaler value of which can be set independently.

14 . The system of claim 9 , wherein the plurality of datasets includes historical data is data collected within a predetermined time window.

15 . The system of claim 14 , wherein the predetermined time window is from the date defined as 1 day previous to the current date to the date defined as 14 days previous to the current date.

16 . A non-transitory computer readable storage medium, including instructions stored thereon for frequent machine learning model retraining and rule optimization in a payment transaction fraud detection system, which when read and executed by one or more computers cause the one or more computers to perform steps comprising:

generating a challenger machine learning model based on a production machine learning model;

training the challenger machine learning model on a plurality of datasets, wherein each of the plurality of datasets includes data records collected on a date defined as a number of days previous to a current date, wherein the plurality of datasets include historical payment transaction data;

scoring the historical payment transaction data with the challenger machine learning model, wherein the scoring produces a respective score for each record of a plurality of records in the plurality of datasets, wherein the scoring produces respective fraud scores for said each record of the plurality of records;

determining that the challenger model performs within predetermined thresholds based on the scoring;

selecting an optimal scaler value for a rule based on execution of the rule with a range of scaler values applied to the respective score for said each record of the plurality of records evaluated by the rule;

determining that the optimal scaler value outperforms a production scaler value; and

promoting the challenger model and the optimal scaler value to a production environment to automatically process payment transaction data for fraud detection by (1) calculating a model score of the payment transaction data and assigning a return on investment value to the payment transaction data based on variables of the payment transaction data and (2) calculating a fraud score of the payment transaction data.