# US 2019/0378049 A1 · Ensemble of machine learning engines coupled to a graph structure that spreads heat

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

- Number: US 2019/0378049 A1
- Document ID: US20190378049A1
- Status: Published application
- Kind: A1
- Application number: 16/006,285
- Filed: 2018-06-12
- Published: 2019-12-12
- Earliest claimed: 2018-06-12
- Assignees: Bank of America Corporation
- Applicants: Bank of America Corporation
- Inventors: Craig Douglas Widmann; Margaret A. Payne; Joel Filliben; Eren Kursun; Lorraine C. Edkin; Dana M. Pusey-Conlin; Ronnie J. Morris; Scott A. Sims
- Web page: https://patentgrep.com/patent/US20190378049A1

## CPC

- G06N 20/00: Machine learning
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/20: Ensemble learning
- G06Q 20/4016: {involving fraud or risk level assessment in transaction processing}
- G06Q 30/0185: {Product, service or business identity fraud}
- G06N 3/048: Activation functions
- G06N 5/025: Extracting rules from data
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- 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 machine learning system that optimizes a feature vector, the system comprising:
a first interface configured to receive transaction data;
a graph module configured to store and update a graph using the 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; and
a machine learning engine comprising a plurality of machine learning sub-engines, wherein each entity type in the graph is assigned a separate machine learning sub-engine, the machine learning engine is programmed to perform steps comprising:
training a machine learning model of a machine learning sub-engine of the machine learning engine using the transaction data;
classifying a plurality of nodes in the graph based on known patterns in the transaction data and the machine learning model, by setting a classification attribute of each node to one of a plurality of classifications;
detecting, by the machine learning sub-engine, an emerging pattern between a first node and second node in the graph based on the transaction data;
inserting an edge between the first node and the second node in the graph in response to the detecting of the emerging pattern; and
adjusting the feature vector based on an objective function to minimize a loss function.

### Claim 2 (depends on 1)

The system of claim 1, wherein the machine learning engine comprising a plurality of machine learning sub-engines is an ensemble, wherein the first node is a first entity type and the second node is not the first entity type, and wherein the machine learning sub-engine is assigned to the first node and is different than a machine learning sub-engine assigned to the second node.

### Claim 3 (depends on 1)

The system of claim 1, wherein the classifying of the first node in the graph comprises populating a confidence attribute of the first node, based on the machine learning model.

### Claim 4 (depends on 3)

The system of claim 3, wherein the machine learning model corresponds to the machine learning sub-engine assigned to the entity type of the first node.

### Claim 5 (depends on 1)

The system of claim 1, wherein the adjusting the feature vector comprises adding a feature to the feature vector.

### Claim 6 (depends on 5)

The system of claim 5, wherein the feature is a hardware identifier assigned to a device originating the transaction data received by the first interface.

### Claim 7 (depends on 5)

The system of claim 5, wherein the feature is one of: a phone number corresponding to a device originating the transaction data received by the first interface, a unique identifier assigned to a cookie corresponding to the transaction data, an email address, and a screen resolution of the device.

### Claim 8 (depends on 1)

The system of claim 1, wherein the adjusting the feature vector comprises removing a feature from the feature vector.

### Claim 9 (depends on 1)

The system of claim 1, the system further comprising:
a user computing device configured to originate the transaction data received by the first interface, wherein the transaction data comprises a hardware identifier assigned to the user computing device, a phone number corresponding to the user computing device, a unique identifier assigned to a cookie corresponding to the transaction data, an email address, and a screen resolution of the user computing device.

### Claim 10 (depends on 1)

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

### Claim 11 (depends on 1)

The system of claim 1, wherein the graph module, in response to receiving a current event data, is configured to set the classification attribute of the first node to a value predicted by the machine learning engine.

### Claim 12 (depends on 11)

The system of claim 11, wherein the current event data comprises at least one of: new transaction data, a report of a stolen card, an uninstallation of a software application from a device, and an installation of a software application onto a device.

### Claim 13 (depends on 1)

The system of claim 1, wherein the graph module is configured to update a confidence attribute of the first node based on the machine learning model that detected the emerging pattern.

### Claim 14 (depends on 2)

The system of claim 2, wherein the ensemble comprises a graphics processing unit.

### Claim 15 (depends on 2)

The system of claim 2, wherein the ensemble comprises a semi-supervised machine learning engine.

### Claim 16 (depends on 2)

The system of claim 2, wherein the first node stored by the graph module corresponds to a device entity type and the second node stored by the graph module corresponds to a customer entity type, and wherein the edge inserted between the first node and the second node is in response to the ensemble detecting the emerging pattern between a customer's smartphone device and fraud.

### Claim 17

An apparatus comprising:
a first interface configured to receive transaction data;
a graph module configured to store and update a graph using the 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; and
an ensemble of machine learning sub-engines programmed to perform steps comprising:
training a machine learning model of a machine learning sub-engine of the ensemble using a corpus, wherein the corpus comprises a training data and a test data;
classifying a plurality of nodes in the graph based on the machine learning model, by setting a classification attribute of a first node and a second node of the plurality of nodes to one of a plurality of classifications; and
inserting an edge in the graph between the first node and the second node in response to the machine learning model detecting a pattern.

### Claim 18 (depends on 17)

The apparatus of claim 17, wherein each entity type in the graph is assigned a separate machine learning sub-engine, and wherein a first machine learning sub-engine of the ensemble of machine learning sub-engines is a neural network, and wherein a second machine learning sub-engine of the ensemble of machine learning sub-engines is a Boltzmann machine, and wherein a third machine learning sub-engine of the ensemble of machine learning sub-engines is a restricted Boltzmann machine, and wherein a fourth machine learning sub-engine of the ensemble of machine learning sub-engines is an autoencoder.

### Claim 19

A non-transitory computer readable medium storing computer-executable instructions that, when executed by a graphics processing unit, cause an ensemble of machine learning sub-engines to:
train a machine learning model of the ensemble of machine learning sub-engines using a corpus, wherein the corpus comprises a training data and a test data;
classify a plurality of nodes in a graph, which comprises nodes and edges and is stored in computer memory, based on the machine learning model, by setting a classification attribute of a first node and a second node of the plurality of nodes to one of a plurality of classifications; and
insert an edge in the graph between the first node and the second node in response to the machine learning model detecting a pattern, wherein the first node corresponds to a first entity type and the second node does not correspond to a second entity type.

### Claim 20 (depends on 19)

The non-transitory computer readable medium of claim 19, wherein the first entity type is a device entity type and the second entity type is a customer entity type, and wherein the edge inserted linking the first node to the second node is in response to the ensemble detecting a smartphone device corresponding to the first node is associated with a fraudulent transaction reported by a customer corresponding to the second node.
