# US 2019/0377819 A1 · Machine learning system to detect, label, and spread heat in a graph structure

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0377819 A1
- Document ID: US20190377819A1
- Status: Published application
- Kind: A1
- Application number: 16/006,388
- Filed: 2018-06-12
- Published: 2019-12-12
- Earliest claimed: 2018-06-12
- Assignees: Bank of America Corporation
- Applicants: Bank of America Corporation
- Inventors: Joel Filliben; Eren Kursun; Lorraine C. Edkin; Scott A. Sims; Craig Douglas Widmann; Margaret A. Payne; Ronnie J. Morris; Dana M. Pusey-Conlin
- Web page: https://patentgrep.com/patent/US20190377819A1

## CPC

- G06F 16/26: Visual data mining; Browsing structured data
- G06N 20/20: Ensemble learning
- G06F 16/23: Updating
- G06F 16/288: {Entity relationship models}
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/091: Active learning
- G06N 3/092: Reinforcement learning
- G06N 20/00: Machine learning
- G06Q 20/4016: {involving fraud or risk level assessment in transaction processing}
- G06N 3/047: Probabilistic or stochastic networks
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

Machine learning models, semantic networks, adaptive systems, artificial neural networks, convolutional neural networks, and other forms of knowledge processing systems are disclosed. An ensemble machine learning system is coupled to a graph module storing a graph structure, wherein a collection of entities and the relationships between those entities forms nodes and connection arcs between the various nodes. A hotfile module and hotfile propagation engine coordinate with the graph module or may be subsumed within the graph module, and implement the various hot file functionality generated by the machine learning systems.

## Claims

### Claim 1

A system comprising:
a machine learning engine comprising a machine learning model trained using historical transaction data to identify a known pattern;
a graph module configured to store and update a graph with incoming transaction data, the graph comprising nodes and edges, wherein each node corresponds to an entity type, and wherein each edge represents a relationship between two nodes;
a hotfile module configured to assign one or more functions to each node in the graph; and
a hotfile propagation engine programmed to perform functions comprising:
detecting in the incoming transaction data the known pattern identified by the machine learning engine;
labeling the incoming transaction data by integrating a first node corresponding to the incoming transaction data into the graph and by inserting an edge linking the first node with an existing node in the graph, wherein the first node is an entity type based on the incoming transaction data;
spreading heat from the existing node to the first node, wherein the heat corresponds to a classification attribute, and wherein an amount of the heat spread is based on a weight assigned to the edge connecting the first node with the existing node;
fading the heat of the first node based on one or more conditions; and
restricting one or more nodes in the graph based on the heat of the first node.

### Claim 2 (depends on 1)

The system of claim 1, further comprising a feature vector, wherein the machine learning engine is configured to minimize a loss function based on an objective function using the feature vector.

### Claim 3 (depends on 2)

The system of claim 2, wherein the feature vector comprises one or more of:
a hardware identifier of a computing device that originated the incoming transaction data,
a phone number of the computing device that originated the incoming transaction data,
a unique identifier assigned to a cookie corresponding to the incoming transaction data,
an email address,
an email domain of the email address,
a screen resolution of the computing device that originated the incoming transaction data, and
an IP address of the computing device that originated the incoming transaction data.

### Claim 4 (depends on 2)

The system of claim 2, wherein the machine learning engine is configured to adjust the feature vector by adding a feature to the feature vector and removing a feature from the feature vector.

### Claim 5 (depends on 1)

The system of claim 1, wherein the machine learning engine is programmed to perform steps comprising:
detecting an emerging pattern between the first node in the graph and fraud;
causing the hotfile module to increase the weight assigned to the edge of the first node, wherein the amount of the heat spread is based on the weight assigned to the edge; and
adjusting a feature vector based on the emerging pattern.

### Claim 6 (depends on 5)

The system of claim 5, wherein the first node corresponds to a customer entity type and the existing node in the graph corresponds to a device entity type.

### Claim 7 (depends on 1)

The system of claim 1, wherein the first node comprises a confidence attribute, and the hotfile module is configured to update the confidence attribute of the first node based on the machine learning model.

### Claim 8 (depends on 1)

The system of claim 1, wherein the labeling function comprises updating a confidence attribute of the first node based on the machine learning model.

### Claim 9 (depends on 1)

The system of claim 1, wherein the heat of the first node is one of: low, medium, and high.

### Claim 10 (depends on 1)

The system of claim 1, wherein the one or more conditions of the fading function comprise time and speed.

### Claim 11 (depends on 1)

The system of claim 1, further comprising:
a historical data store communicatively coupled to the graph module, wherein the historical data store contains the historical transaction data.

### Claim 12 (depends on 1)

The system of claim 1, further comprising:
a user computing device that originated the incoming transaction data, wherein the user computing device is configured to receive user input through a graphical user interface of the user computing device then transmit the user input to the machine learning system;
wherein the machine learning system comprises a supervised learning system that trains the machine learning model using the user input.

### Claim 13 (depends on 1)

The system of claim 1, wherein the machine learning engine comprises a hardware graphics processing unit configured to execute train the machine learning model.

### Claim 14

A method to automatically detect, label, spread, and fade heat in a graph stored in a graph module, wherein the graph module is communicatively coupled to a machine learning engine, a hotfile module, and a hotfile propagation engine, the method comprising:
detecting in an incoming transaction data a known pattern identified by the machine learning engine;
labeling the incoming transaction data by integrating a first node corresponding to the incoming transaction data into the graph and by inserting an edge linking the first node with an existing node in the graph, wherein the first node is an entity type based on the incoming transaction data;
spreading heat from the existing node to the first node, wherein the heat corresponds to a classification attribute, and wherein an amount of the heat spread is based on a weight assigned to the edge connecting the first node with the existing node; and
restricting one or more nodes in the graph based on the heat of the first node.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
fading the heat of the first node based on one or more conditions.

### Claim 16 (depends on 15)

The method of claim 15, wherein the one or more conditions of the fading comprises time and speed.

### Claim 17 (depends on 14)

The method of claim 14, wherein the heat of the first node is one of: low, medium, and high.

### Claim 18 (depends on 14)

The method of claim 14, wherein the heat is a numeric value within a predetermined range.

### Claim 19

A computer memory storing computer-executable instructions that, when executed by a graphics processing unit (GPU), cause a graph module, which is communicatively coupled to a machine learning engine, to:
detect, by the GPU, in an incoming transaction data a known pattern identified by the machine learning engine;
label, by the GPU, the incoming transaction data by integrating a first node corresponding to the incoming transaction data into a graph and by inserting an edge linking the first node with an existing node in the graph, wherein the first node is an entity type based on the incoming transaction data, and wherein the graph comprising nodes and edges, wherein each node corresponds to an entity type, and wherein each edge represents a relationship between two nodes;
spread heat, by the GPU, from the existing node to the first node, wherein the heat corresponds to a classification attribute, and wherein an amount of the heat spread is based on a weight assigned to the edge connecting the first node with the existing node;
fade, by the GPU, the heat of the first node based on one or more conditions; and
restrict one or more nodes in the graph based on the heat of the first node.

### Claim 20 (depends on 19)

The computer memory of claim 19, wherein the machine learning engine comprises a machine learning model trained using historical transaction data, and wherein the heat is a numeric value within a predetermined range.
