Granted Patent B2

US 12,596,963 · App. 17/858,581 · Granted Apr 7, 2026

Inventors: Gunther Havel (Portsmouth, NH); John Wilson (Richmond, VA); Ashwin Assysh Sharma (Minneapolis, MN)

Assignee: Capital One Services, LLC

G06N20/20G06F18/2113G06F18/22G06Q20/389

Abstract

Claims (75)

1 . A method comprising:

receiving, by a computing device from a cloud storage device, a trigger command indicating a threshold is satisfied by a quantity of records that have been uploaded from a first database;

receiving, by the computing device based on the trigger command, the uploaded records comprising a plurality of source records, from a first database, and a plurality of target records from a second database, wherein the first database is configured to store records of a different format from records in the second database;

inputting, based on the trigger command and into a machine learning model, the plurality of source records and the plurality of target records;

receiving, as output from the machine learning model, a selection of a subset of the plurality of source records and a subset of the plurality of target records, wherein the selection is based on:

determining, for each of the plurality of target records, a distance value between:

the respective target record, and

the subset of the plurality of source records;

ranking, based on the distance value of each of the plurality of target records, the plurality of target records;

selecting, based on the ranking, the subset of target records of the plurality of target records;

generating a matched record that indicates a matching between:

the selected s subset of the plurality of target records, and

the selected subset of the plurality of source records;

sending, to the first database, the matched record; and

updating, based on the matched record, the first database by changing a status of the subset of the plurality of source records from unmatched to matched.

2 . The method of claim 1 , wherein each of the plurality of source records and the plurality of target records is a transaction record that indicates a balance amount, and wherein generating the matched record is further based on determining that a total balance amount of the subset of the plurality of source records equals to a total balance amount of the selected one or more target records.

3 . The method of claim 1 , wherein the receiving the plurality of source records comprises receiving, from the first database via a cloud storage bucket, the plurality of source records.

4 . The method of claim 3 , wherein the plurality of source records and the plurality of target records are unmatched before being input into the machine learning model.

5 . The method of claim 1 , wherein each of the plurality of target records comprises a plurality of data fields, and wherein determining the distance value comprises:

identifying one or more data fields of the plurality of data fields; and

calculating the distance value based on data within the one or more data fields.

6 . The method of claim 1 , wherein the distance value is a gower distance value.

7 . The method of claim 1 , further comprising:

dividing, based on a second machine learning model comprising a clustering algorithm, the plurality of source records into one or more subsets.

8 . The method of claim 1 , wherein selecting the subset of the plurality of source records comprises determining a distance value between each two source records of the subset does not exceed a threshold.

9 . A system comprising:

a computing device, and

a first database;

wherein the computing device comprising:

one or more processors;

instructions, when executed by the one or more processors, cause the computing device to:

receive, from a cloud storage device, a trigger command indicating a threshold is satisfied by a quantity of records that have been uploaded from a first database;

receive, based on the trigger command, the uploaded records comprising, a plurality of source records, from a first database, and a plurality of target records from a second database, wherein the first database is configured to store records of a different format from records in the second database;

input, based on the trigger command and into a machine learning model, the plurality of source records and the plurality of target records;

receive, as output from the machine learning model, a selection of a subset of the plurality of source records and a subset of the plurality of target records, wherein the selection is based on:

determine, for each of the plurality of target records, a distance value between:

the respective target record, and

the subset of the plurality of source records;

rank, based on the distance value of each of the plurality of target records, the plurality of target records;

select, based on the ranking, the subset of the plurality of target records;

generate a matched record that indicates a matching between:

the selected subset of the plurality of target records, and

the selected subset of the plurality of source records;

send, to the first database, the matched record; and

update, based on the matched record, the first database by changing a status of the subset of the plurality of source records from unmatched to matched; and

wherein the first database is configured to receive, from the computing device, the matched record.

10 . The system of claim 9 , wherein each of the plurality of source records and the plurality of target records is a transaction record that indicates a balance amount, and wherein the computing device is configured to generate the matched record further based on determining that a total balance amount of the subset of the plurality of source records equals to a total balance amount of the selected one or more target records.

11 . The system of claim 9 , wherein the instructions, when executed by the one or more processors, cause the computing device to receive the plurality of source records from the first database via a cloud bucket.

12 . The system of claim 11 , wherein the plurality of source records and the plurality of target records are unmatched before being input into the machine learning model.

13 . The system of claim 9 , wherein each of the plurality of target records comprises a plurality of data fields, and wherein the instructions, when executed by the one or more processors, cause the computing device to determine the distance value by:

identifying one or more data fields of the plurality of data fields; and

calculating the distance value based on data within the one or more data fields.

14 . The system of claim 9 , wherein the distance value is a gower distance value.

15 . The system of claim 9 , wherein the instructions, when executed by the one or more processors, further cause the computing device to:

divide, based on a second machine learning model comprising a clustering algorithm, the plurality of source records into one or more subsets.

16 . The system of claim 9 , wherein the instructions, when executed by the one or more processors, cause the computing device to select the subset of the plurality of source records by determining a distance value between each two source records of the subset does not exceed a threshold.

17 . A non-transitory computer-readable medium storing computer instruction that, when executed by one or more processors, cause performance of actions comprising:

receiving, by a computing device, from a cloud storage device, a trigger command indicating a threshold is satisfied by a quantity of records that have been uploaded from a first database;

receiving, by the computing device based on the trigger command, the uploaded records comprising a plurality of source records, from a first database, and a plurality of target records from a second database, wherein the first database is configured to store records of a different format from records in the second database;

inputting, based on the trigger command and into a machine learning model, the plurality of source records and the plurality of target records;

receiving, as output from the machine learning model and based on a clustering algorithm, a selection of a subset of the plurality of source records and a subset of the plurality of target records, wherein the selection is based on:

determining a distance value between each two source records of the subset does not exceed a threshold;

determining, based on a second machine learning model and for each of the plurality of target records, a distance value between:

the respective target record, and

the subset of the plurality of source records;

ranking, based on the distance value of each of the plurality of target records, the plurality of target records;

selecting, based on the ranking, the subset of the plurality of target records;

generating a matched record that indicates a matching between:

the selected subset of the plurality of target records, and

the selected subset of the plurality of source records;

sending, to the first database, the matched record; and

updating, based on the matched record, the first database by changing a status of the subset of the plurality of source records from unmatched to matched.

18 . The non-transitory computer-readable medium of claim 17 , wherein each of the plurality of source records and the plurality of target records is a transaction record that indicates a balance amount, and wherein generating the matched record is further based on determining that a total balance amount of the subset of the plurality of source records equals to a total balance amount of the selected one or more target records.

19 . The non-transitory computer-readable medium of claim 17 , wherein the instructions, when executed by the one or more processor, cause receiving the plurality of source records from the first database via a cloud storage bucket.

20 . The non-transitory computer-readable medium of claim 19 , wherein the plurality of source records and the plurality of target records are unmatched before being input into the machine learning model.

Assignments (1)

Continuity (1)

References Cited (17)