Network based features for financial crime detection

Patent metadata

Field	Value
Publication number	US-11640609-B1
Application number	US-201916714027-A
Country	US
Kind code	B1
Filing date	Dec 13, 2019
Priority date	Dec 13, 2019
Publication date	May 2, 2023
Grant date	May 2, 2023

Abstract

Disclosed is an example approach in which network and non-network features are used to train a predictive machine learning model that is implemented to predict financial crime and fraud. Graphical network features may be generated by applying financial entity risk vectors to a network model with representations of various types of networks. The network model may comprise transactional, non-social, and/or social networks, with edges corresponding to linkages that may be weighted according to various characteristics (such as frequency, amount, type, recency, etc.). The graphical network features may be fed to the predictive model to generate a likelihood and/or prediction with respect to a financial crime. A perceptible alert is generated on one or more computing devices if a financial crime is predicted or deemed sufficiently likely. The alert may identify a subset of the set of financial entities involved in the financial crime and present graphical representations of networks and linkages.

First claim

What is claimed is:

1. A machine learning method implemented by a computing system of a financial institution, the method comprising:
training, by the computing system, a machine learning predictive model using historic graphical network features as training data, the machine learning predictive model trained to generate predictions of financial crimes and perceptible alerts identifying financial entities involved in financial crime on one or more computing devices in response to predictions of financial crimes;
generating, by the computing system, based on a plurality of data records accessed via an electronic database, a network model including a set of financial entities and relationships therebetween, generation of the network model comprising generating an adjacency matrix;
retrieving, by the computing system, from the electronic database, a plurality of financial entity risk vectors monitored by the financial institution;
generating, by the computing system, graphical network features by multiplying financial entity risk vectors by the adjacency matrix of the network model, the adjacency matrix being a function of discrete time, the financial entity risk vectors multiplied by the adjacency matrix according to a plurality of propagation steps;
executing, by the computing system, the machine learning predictive model using the graphical network features as input to generate a prediction of a financial crime, the machine learning predictive model comprising a density-based clustering technique that is a function of a density parameter; and
generating, by the computing system, a perceptible alert on one or more computing devices in response to the prediction of the financial crime, the alert identifying a subset of the set of financial entities involved in the financial crime, the alert comprising an (i) interactive scroll element that enables presentation of additional elements of the subset of the set of financial entities, and (ii) an interactive update button that causes the alert to be updated with subsequent data corresponding to the subset of the set of financial entities.

2. The machine learning method of claim 1 , wherein the network model comprises a financial transaction network of financial transactions between financial entities in the set of financial entities.

3. The machine learning method of claim 2 , wherein the computing system defines each edge in the financial transaction network according to at least one of an amount, a frequency, or an incidence of financial transactions between a corresponding pair of financial entities in the set of financial entities.

4. The machine learning method of claim 3 , wherein generating the network model comprises detecting, by the computing system, financial transactions in an internal transaction database of the financial institution.

5. The machine learning method of claim 1 , wherein the network model comprises a non-social relationship network of non-social relationships between financial entities in the set of financial entities.

6. The machine learning method of claim 5 , wherein the computing system defines each edge in the non-social relationship network as denoting existence of a non-social relationship between a corresponding pair of financial entities, each edge weighted according to a characteristic of the corresponding non-social relationship.

7. The machine learning method of claim 6 , wherein generating the network model comprises detecting, by the computing system, non-social relationships using customer data stored in a customer database of the financial institution.

8. The machine learning method of claim 1 , wherein the network model comprises a social network corresponding to social linkages between financial entities.

9. The machine learning method of claim 8 , wherein the computing system defines each edge in the social network as denoting a social linkage between corresponding pairs of financial entities, each edge weighted according to a characteristic of the corresponding social linkage.

10. The machine learning method of claim 9 , wherein generating the network model comprises detecting, via the computing system, social linkages using data accessed via at least one of: social media devices; or internet-accessible devices that store news reports or biographies of financial entities.

11. The machine learning method of claim 1 , wherein the risk vectors are real-value risk vectors comprising at least one of: transaction-based risk vectors; or customer risk scores.

12. The machine learning method of claim 11 , wherein the network model comprises adjacency matrices, and wherein generating the graphical network features comprises generating, by the computing system, graphical network features for combinations of adjacency matrices and risk vectors compatible with corresponding adjacency matrices.

13. A machine learning computing system of a financial institution, the computing system comprising a processor and a memory comprising instructions executable by the processor, the instructions comprising a machine learning platform configured to:
training, by the computing system, a machine learning predictive model using historic graphical network features as training data, the machine learning predictive model trained to generate predictions of financial crimes and perceptible alerts identifying financial entities involved in financial crime on one or more computing devices in response to predictions of financial crimes;
generating, based on a plurality of data records, a network model including a set of financial entities and relationships therebetween;
generate graphical network features by multiplying financial entity risk vectors by an adjacency matrix of the network model, the adjacency matrix being a function of discrete time, the financial entity risk vectors multiplied by the adjacency matrix according to a plurality of propagation steps;
execute the machine learning predictive model using the graphical network features as input to generate a prediction of a financial crime, the machine learning predictive model comprising a density-based clustering technique that is a function of a density parameter; and
generate a perceptible alert on one or more computing devices in response to the prediction of the financial crime, the alert identifying a subset of the set of financial entities involved in the financial crime, the alert comprising (i) an interactive scroll element that enables presentation of additional elements of the subset of the set of financial entities, and (ii) an interactive update button that causes the alert to be updated with subsequent data corresponding to the subset of the set of financial entities.

14. The machine learning system of claim 13 , wherein the network model comprises a financial transaction network of financial transactions between financial entities in the set of financial entities, and wherein the machine learning platform defines each edge in the financial transaction network according to at least one of an amount, a frequency, or an incidence of financial transactions between a corresponding pair of financial entities in the set of financial entities.

15. The machine learning system of claim 13 , wherein the network model comprises a non-social relationship network of non-social relationships between financial entities in the set of financial entities, and wherein the machine learning platform defines each edge in the non-social relationship network as denoting existence of a non-social relationship between a corresponding pair of financial entities, each edge weighted according to a characteristic of the corre

Assignees

• Wells Fargo Bank Na

Inventors

• Shoumaker Wayne B
• Singhal Harsh
• Sreehari Suhas
• Sudjianto Agus
• Yu Ye

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06Q40/02 Primary

  Banking, e.g. interest calculation or account maintenance (credit or loans G06Q40/03) · CPC title

• G06Q40/00

  Finance; Insurance; Tax strategies; Processing of corporate or income taxes · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N20/00

  Machine learning · CPC title