System for detecting associated records in a record log

Patent metadata

Field	Value
Publication number	US-12159287-B2
Application number	US-202117248460-A
Country	US
Kind code	B2
Filing date	Jan 26, 2021
Priority date	Jan 26, 2021
Publication date	Dec 3, 2024
Grant date	Dec 3, 2024

Abstract

In some implementations, a system may identify a mitigation record in a record log associated with an account of an application. The system may append a message prompt to the mitigation record that causes the application to request, during a user session associated with the account, an indication of whether there is an association between the mitigation record and an event record. The system may receive, from a user device associated with the user session, feedback associated with the message prompt that indicates whether the mitigation record is associated with the event record. The system may perform, based on the feedback, an action associated with indicating whether the mitigation record and the event record are associated.

First claim

What is claimed is:

1. A system for detecting associated records in a record log, the system comprising:
one or more memories; and
one or more processors, communicatively coupled to the one or more memories, configured to:
monitor the record log of an account associated with an application;
identify a mitigation record in the record log;
analyze, based on identifying the mitigation record, the record log to identify a subset of event records in the record log that are associated with one or more return attributes of the mitigation record;
train, based on using a set of observations and one or more machine learning algorithms, an association model,
wherein the one or more machine learning algorithms include at least one of a regression algorithm, a decision tree algorithm, a neural network algorithm, a k-nearest neighbor algorithm, and a support vector machine algorithm;

generate, based on applying the trained association model to a new observation that includes at least one of a time period, a second feature of entity identifiers, or a third feature of a value sum, an output that includes at least one of:
a predicted value of a target variable when supervised learning is utilized,
information that identifies a cluster to which the new observation belongs when unsupervised learning is utilized, or
information that indicates a degree of similarity between the new observation and other observations;

determine, based on the output, an association score associated with the mitigation record and a particular event record of the subset of event records,
wherein the association score is indicative of a probability that the mitigation record and the particular event record are associated with a same event;

determine that the association score indicates that the probability satisfies a probability threshold;
append and store, to the mitigation record and based on determining that the association score indicates that the probability satisfies the probability threshold, a message prompt configured to be displayed during a session with a user device, that causes the application to indicate an association between the mitigation record and the particular event record when the particular event record is presented via a display of the user device that is executing the application,
wherein information associated with the message prompt indicates that the message prompt is to be displayed during the session;

monitor the record log being accessed by the user device;
determine, based on the monitoring, that the particular event record is being accessed by the user device and is associated with the message prompt;
cause display of the appended and stored message prompt based on determining that the particular event record is being accessed by the user device;
receive, from the user device, feedback associated with the message prompt that verifies whether the mitigation record is associated with the event record;
retrain, based on the received feedback, the association model;
perform an action associated with the feedback, the mitigation record, and the event record;
use the retrained association model for subsequent processing to determine another association score associated with another mitigation record; and
remove, based on performing the action, the message prompt by replacing the message prompt with information indicating the association between the mitigation record and the particular event record.

2. The system of claim 1 , wherein the one or more processors, when analyzing the record log, are configured to at least one of:
filter, from records of the record log, other mitigation records from the record log;
filter, from records of the record log, certain event records that are outside of a time period associated with a time of the mitigation record that is identified by a timestamp attribute of the one or more return attributes;
filter, from records of the record log, certain event records that are associated with a value that is outside of a range from zero to an additive inverse of a value attribute of the mitigation record that is identified by the one or more return attributes; or
filter, from records of the record log, certain event records that are not associated with an entity that is identified by an entity attribute of the one or more return attributes.

3. The system of claim 1 , wherein the probability threshold is based on the association score.

4. The system of claim 1 , wherein the association model comprises a machine learning model that is trained to identify associations between newly identified mitigation records and previously stored event records of a plurality of accounts of the application,
wherein the machine learning model is trained based on historical records associated with the plurality of accounts and previously determined associations between historical mitigation records of the historical records and historical event records of the historical records.

5. The system of claim 1 , wherein the one or more processors, when performing the action, are configured to:
determine that the feedback includes a verification that the mitigation record is associated with the particular event record; and
replace the message prompt with an indication that the mitigation record is associated with the particular event record.

6. The system of claim 5 , wherein the one or more processors are further configured to:
append, to the event record, another indication that the mitigation record is associated with the particular event record.

7. The system of claim 1 , wherein the one or more processors, when performing the action, are configured to:
determine that the feedback includes an indication that the mitigation record is not associated with the particular event record;
remove the message prompt from the mitigation record; and
update the association model based on the feedback, the mitigation record, and the particular event record.

8. A method for detecting associated records in a record log, comprising:
identifying, by a device, a mitigation record in the record log associated with an account of an application;
training, based on using a set of observations that do not include a target variable and one or more machine learning algorithms, an association model;
generating, based on applying the trained association model to a new observation that includes at least one of a time period, a second feature of entity identifiers, or a third feature of a value sum, an output;
determining, by the device and based on the output, an association score associated with the mitigation record and a particular event record of the record log,
wherein the association score is indicative of a probability that the mitigation record and the particular event record are associated with a same event, and
wherein the association score is determined based on a comparison of return attributes of the mitigation record and charge attributes of the particular event record;

determining, by the device, that the association score indicates that the probability satisfies a probability threshold;
appending and storing, by the device and based on determining that the associated score indicates that the probability satisfies the probability threshold, a message prompt to the mitigation record,
wherein the message prompt is configured to be displayed during a session with a user device, and
wherein the message prompt causes the application to request, during the session, an indication of whether there is an association between the mitigation record and the particular event record;

monitoring, by the device, the record log being accessed by the user device;
determining, by the device and based on the monitoring, that the particu

Assignees

• Capital One Services Llc

Inventors

• Lee Jonathan
• Moreau-Arnott Elijah
• Frensch Steve
• Randall Garrett

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/091

  Active learning · CPC title

• G06Q40/12

  Accounting · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title

• G06Q20/10

  specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems · CPC title