Time-based input and output monitoring and analysis to predict future inputs and outputs

Patent metadata

Field	Value
Publication number	US-12014430-B2
Application number	US-202217661304-A
Country	US
Kind code	B2
Filing date	Apr 29, 2022
Priority date	Apr 29, 2022
Publication date	Jun 18, 2024
Grant date	Jun 18, 2024

Abstract

A monitoring system for detecting anomalous discharges stores input and output records, and fetches and discharges corresponding quantities to and from user resources. The output event records are discriminated for category attributes, and user-specific output category-specific metrics are aggregated for comparison to peer-representative output-category indexes. An alert is sent across a network connection for display on a user device upon determining the user-specific output category-specific metric diverges from the peer-representative output-category index. An alert, in some examples, indicates that the user-specific output category metric exceeds the peer-representative output-category index. Exposure of user-entity resources to outputs that are divergent from peer-representative levels is mitigated. The alert is advantageous to user entities and user devices, enabling early action to be taken by the user entity.

First claim

What is claimed is:

1. A monitoring system for detecting anomalous discharges, the monitoring system comprising:
a computing system including one or more processor and at least one of a memory device and a non-transitory storage device, wherein said one or more processor executes computer-readable instructions; and
a network connection operatively connecting user devices to the computing system,
wherein, upon execution of the computer-readable instructions, the computing system performs steps comprising, for each specific user entity of multiple user entities:
storing input event records associated with the specific user entity, each of the input event records representing a respective quantized input event;
fetching, for at least some of the input event records, a respective input quantity to one or more resource of the specific user entity;
storing output event records associated with the specific user entity, each of the output event records representing a respective quantized output event;
discharging, for at least some of the output event records, a respective output quantity from the one or more resource of the specific user entity;
discriminating, for at least some of the output event records, a respective at least one output category-specific attribute;
associating each output event record, for which a respective at least one output category-specific attribute is discriminated, with each corresponding respective output category;
aggregating over a time interval, for each output category with which output event records of the specific user-entity are associated, a user-specific output category metric;
training, via machine learning and using a set of training data, an aggregating algorithm to identify an aggregation of peer entities to respective user entities and generate a peer-representative output-category index, the training including:
iteratively predicting which of multiple user entities should be included in the aggregation of peer entities of a respective user entity, the predicting being based on at least one output category with which output event records of multiple user-entities are associated;
testing and comparing the aggregation of peer entities predicted during each iteration against a target variable; and
indicating, via a feedback loop, for each iteration whether modifications to weights assigned to certain entity data of the multiple user entities are necessary to improve predictability of the target variable;

deploying the trained aggregating algorithm to generate a peer-representative output-category index for the specific user-entity using one or more user-specific output category metrics and based thereon determining, for each output category for which the user-specific output category metric is aggregated, whether the user-specific output category metric diverges from the peer-representative output-category index for the specific user-entity; and
transmitting a control signal across the network connection to at least one user device associated with the specific user entity upon determining the user-specific output category metric diverges from the peer-representative output-category index, the control signal initiating display of an indication that the user-specific output category metric diverges from the peer-representative output-category index by exceeding the peer-representative output-category index.

2. The monitoring system of claim 1 , the steps further comprising aggregating over time, via the trained aggregating algorithm, for at least one output category with which output event records of multiple user-entities are associated, the peer-representative output-category index for the specific user-entity.

3. The monitoring system of claim 2 , wherein the trained aggregating algorithm aggregates the peer-representative output-category index for the specific user-entity from the output event records of peer entities of the specific user-entity from at least one time period preceding said time interval.

4. The monitoring system of claim 1 , wherein discriminating the respective at least one output category-specific attribute comprises using a discriminating algorithm trained by a machine-learning technique.

5. The monitoring system of claim 4 , wherein the machine-learning technique utilizes output event records of the multiple user-entities from at least one time period preceding said time interval to train the discriminating algorithm to discriminate category-specific attributes.

6. The monitoring system of claim 4 , wherein the machine-learning technique utilizes output event records of the specific user entity from multiple time periods preceding said time interval to discriminate category-specific attributes using the trained discriminating algorithm.

7. The monitoring system of claim 1 , wherein the peer-representative output-category index represents at least one of: an average; a mean; a normalized sum; and a weighted sum.

8. The monitoring system of claim 1 , the steps further comprising determining the peer-representative output-category index at least in part using third-party data.

9. The monitoring system of claim 1 , wherein the user-specific output category metric represents at least one of: an average; a mean; a normalized sum; and a weighted sum.

10. The monitoring system of claim 1 , wherein the iteratively predicting which of the multiple user entities should be included in the aggregation of peer entities of the respective user entity is specific to a geographical location of the specific user entity.

11. A monitoring system for detecting anomalous discharges, the monitoring system comprising:
a computing system including one or more processor and at least one of a memory device and a non-transitory storage device, wherein said one or more processor executes computer-readable instructions; and
a network connection operatively connecting user devices to the computing system,
wherein, upon execution of the computer-readable instructions, the computing system performs steps for aggregating, for each specific user entity of multiple user entities, output user-specific category-specific output metrics, the steps comprising:
storing input event records associated with the specific user entity, each of the input event records representing a respective quantized input event;
fetching, for at least some of the input event records, a respective input quantity to one or more resource of the specific user entity;
storing output event records associated with the specific user entity, each of the output event records representing a respective quantized output event;
discharging, for at least some of the output event records, a respective output quantity from the one or more resource of the specific user entity;
training, via machine learning and using a set of training data, a discriminating algorithm to discriminate category-specific attributes, the training data comprising output event records of multiple user-entities, the training comprising:
iteratively predicting which category-specific attributes are to be discriminated from the output event records of the multiple user-entities;
testing and comparing predicted category-specific attributes against that are discriminated against a target variable; and
indicating, via a feedback loop, for each iteration whether modifications to weights assigned to certain category-specific attributes are necessary to improve predictability of the target variable;

deploy the trained discriminating algorithm and based thereon discriminating, using the trained discriminating algorithm, for at least some of the output event records, a respective at least one output category-specific attribute;
a

Assignees

• Truist Bank

Inventors

• Green Kevin
• Fitzgerald Raphael
• Mistor Paul Gerard

Classifications

• G06Q40/128 Primary

  Check-book balancing, updating or printing arrangements · CPC title