IP Library › Granted Patent US 12,259,926
Granted Patent B2
US 12,259,926 · App. 18/304,272 · Granted Mar 25, 2025

Computer systems and methods for building and analyzing data graphs

Inventors: Kenrick Fernandes (Wheeling, IL); Ashkan Golgoon (Glenview, IL); Arjun Ravi Kannan (Buffalo Grove, IL)
Assignee: Discover Financial Services
G06F16/9024
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,259,926
App. No.
18/304,272
Granted
Mar 25, 2025
Kind
B2
Abstract

A computing platform may be configured to (i) obtain an input dataset, (ii) construct a graph from the input dataset, (iii) for a given node within the constructed graph, generate a first type of embedding vector using a first embedding technique (e.g., a shallow embedding technique) and a second type of embedding vector using a second embedding technique that differs from the first embedding technique (e.g., a deep embedding technique), and (iv) use the first and second types of embedding vectors for the given node and a data science model to render a given prediction for the given node.

Claims (42)

1. A computing platform comprising:

a network interface;

at least one processor;

at least one non-transitory computer-readable medium; and

program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor such that the computing platform is configured to:

receive, from a client device via the network interface, configuration information for a graph analysis pipeline;

based on the received configuration information, deploy the graph analysis pipeline, wherein the graph analysis pipeline functions to:

obtain an input dataset from either or both of (i) one or more data sources that are internal to the computing platform and (ii) one or more data sources that are external to the computing platform and accessible via one or more networks;

construct a graph from the input dataset;

for a given node within the constructed graph, generate a first type of embedding vector using a first embedding technique and a second type of embedding vector using a second embedding technique that differs from the first embedding technique, wherein the first type is different from the second type; and

input the first and second types of embedding vectors for the given node into a graph data science model that is run on the computing platform and is configured to render a given prediction for the given node by (i) using a first intermediate graph data science model to render a first intermediate prediction based at least in part on the first type of embedding vector, (ii) using a second intermediate graph data science model to render a second intermediate prediction based at least in part on the second type of embedding vector, and (iii) combining the first and second intermediate predictions to produce the given prediction for the given node as an output that comprises a combination of the first and second intermediate predictions, wherein the second intermediate graph data science model is different from the first intermediate graph data science model.

2. The computing platform of claim 1 , wherein the first embedding technique comprises a shallow embedding technique and the second embedding technique comprises a deep embedding technique.

3. The computing platform of claim 1 , wherein one or both of (i) the first intermediate graph data science model renders the first intermediate prediction based further on attribute data for the given node and (ii) the second intermediate graph data science model renders the second intermediate prediction based further on attribute data for the given node.

4. The computing platform of claim 1 , wherein the graph has a configuration that is tailored for a particular use case of the graph.

5. The computing platform of claim 1 , wherein the graph comprises a heterogenous graph that includes two or more different types of nodes.

6. The computing platform of claim 5 , wherein the given prediction for the given nodes comprises a prediction related to financial activity of a given customer of a financial institution, and wherein the two or more different types of nodes comprise two or more of customer nodes, merchant nodes, or transaction nodes.

7. The computing platform of claim 1 , wherein the one or more data sources that are internal to the computing platform comprise one or more internal data stores that contain one or both of raw and pre-processed data.

8. The computing platform of claim 1 , wherein the graph comprises a set of nodes having associated attribute data and a set of edges that connect a respective pair of nodes.

9. The computing platform of claim 1 , wherein the first intermediate graph data science model and the second intermediate graph data science model are trained using a respective machine learning process.

10. The computing platform of claim 1 , wherein combining the first and second intermediate predictions to produce the given prediction for the given node as an output comprises generating a weighted, linear combination of the first and second intermediate predictions.

11. A non-transitory computer-readable medium comprising program instructions that, when executed by at least one processor, cause a computing platform to:

receive, from a client device via a network interface of the computing platform, configuration information for a graph analysis pipeline; and

based on the received configuration information, deploy the graph analysis pipeline, wherein the graph analysis pipeline functions to:

obtain an input dataset from either or both of (i) one or more data sources that are internal to the computing platform and (ii) one or more data sources that are external to the computing platform and accessible via one or more networks;

construct a graph from the input dataset;

for a given node within the constructed graph, generate a first type of embedding vector using a first embedding technique and a second type of embedding vector using a second embedding technique that differs from the first embedding technique, wherein the first type is different from the second type; and

input the first and second types of embedding vectors for the given node into a graph data science model that is run on the computing platform and is configured to render a given prediction for the given node by (i) using a first intermediate graph data science model to render a first intermediate prediction based at least in part on the first type of embedding vector, (ii) using a second intermediate graph data science model to render a second intermediate prediction based at least in part on the second type of embedding vector, and (iii) combining the first and second intermediate predictions to produce the given prediction for the given node as an output that comprises a combination of the first and second intermediate predictions, wherein the second intermediate graph data science model is different from the first intermediate graph data science model.

12. The non-transitory computer-readable medium of claim 11 , wherein the first embedding technique comprises a shallow embedding technique and the second embedding technique comprises a deep embedding technique.

13. The non-transitory computer-readable medium of claim 11 , wherein the graph has a configuration that is tailored for a particular use case of the graph.

14. The non-transitory computer-readable medium of claim 11 , wherein the graph comprises a heterogenous graph that includes two or more different types of nodes.

15. The non-transitory computer-readable medium of claim 14 , wherein the given prediction for the given nodes comprises a prediction related to financial activity of a given customer of a financial institution, and wherein the two or more different types of nodes comprise two or more of customer nodes, merchant nodes, or transaction nodes.

16. The non-transitory computer-readable medium of claim 11 , wherein one or both of (i) the first intermediate graph data science model renders the first intermediate prediction based further on attribute data for the given node and (ii) the second intermediate graph data science model renders the second intermediate prediction based further on attribute data for the given node.

17. A computer-implemented method comprising:

receiving, from a client device, configuration information for a graph analysis pipeline;

based on the received configuration information, deploying, by a computing platform, the graph analysis pipeline, wherein the graph analysis pipeline functions to:

obtain an input dataset from either or both of (i) one or more data sources that are internal to the computing platform and (ii) one or more data sources that are external to the computing platform and accessible via one or more networks;

construct a graph from the input dataset;

for a given node within the constructed graph, generate a first type of embedding vector using a first embedding technique and a second type of embedding vector using a second embedding technique that differs from the first embedding technique, wherein the first type is different from the second type; and

input the first and second types of embedding vectors for the given node into a graph data science model that is run on the computing platform and is configured to render a given prediction for the given node by (i) using a first intermediate graph data science model to render a first intermediate prediction based at least in part on the first type of embedding vector, (ii) using a second intermediate graph data science model to render a second intermediate prediction based at least in part on the second type of embedding vector, and (iii) combining the first and second intermediate predictions to produce the given prediction for the given node as an output that comprises a combination of the first and second intermediate predictions, wherein the second intermediate graph data science model is different from the first intermediate graph data science model.

18. The computer-implemented method of claim 17 , wherein the first embedding technique comprises a shallow embedding technique and the second embedding technique comprises a deep embedding technique.

19. The computer-implemented method of claim 17 , wherein one or both of (i) the first intermediate graph data science model renders the first intermediate prediction based further on attribute data for the given node and (ii) the second intermediate graph data science model renders the second intermediate prediction based further on attribute data for the given node.

20. The computer-implemented method of claim 17 , wherein combining the first and second intermediate predictions to produce the given prediction for the given node as an output comprises generating a weighted, linear combination of the first and second intermediate predictions.

Assignments (2)
MERGER Recorded Jul 2, 2025
From: DISCOVER FINANCIAL SERVICES
To: CAPITAL ONE FINANCIAL CORPORATION
Reel/Frame 071784/0903 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 12, 2023
From: FERNANDES, KENRICK; GOLGOON, ASHKAN; KANNAN, ARJUN RAVI
To: DISCOVER FINANCIAL SERVICES
Reel/Frame 063623/0728 →
Continuity (1)
Related Publication 20240354344A1 · Oct 24, 2024
References Cited (24)
US 20210326389A1 · Sankar et al. · 2021 [cited by applicant]
US 20220286416A1 · Cao et al. · 2022 [cited by applicant]
US 20220318715A1 · Goel et al. · 2022 [cited by applicant]
US 20220382741A1 · Sznajdman · 2022 [cited by examiner]
US 20230162051A1 · Lv · 2023 [cited by examiner]
International Searching Authority. International Search Report and Written Opinion issued in International Application No. PCT/US2024/025706, mailed on Aug. 5, 2024, 10 pages. [cited by applicant]
Svenningsson, Josef et al. Combining Deep and Shallow Embedding for EDSL. Proceedings of the 2012 Conference on Trends in Functional Programming. 2012. 16 pages. [cited by applicant]
Ying et al., GNNExplainer: Generating Explanations for Graph Neural Networks. <URL:https://arxiv.orgpdf/1903.03894v4.pdf>, Nov. 13, 2019, 13 pages. [cited by applicant]
Yuan et al., Explainability in graph neural networks: a taxonomic survey, IEEE Transactions on Pattern Analysis and Machine Intelligence, <URL:https://arxiv.org/pdf/2012.15445.pdf>, Jul. 1, 2022, 19 pages. [cited by applicant]
DeepFindr. How to Explain Graph Neural Networks (with XAI) <URL:https://www.youtube.com/watch?v=NvDM2j8Jgvk>, Oct. 21, 2021, 2 pages. [cited by applicant]
Pope et al., Explainability methods for graph convolutional neural networks, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, <URL: https://openaccess.thecvf.com/content_CVPR_2019/paper… [cited by applicant]
Yuan et al., On explainability of graph neural networks via subgraph explorations, International Conference on Machine <URL:https://arxiv.org/pdf/2102.05152.pdf>, May 31, 2021, pp. 12241-12252. [cited by applicant]
Schnake et al., Higher-Order Explanations of Graph Neural Networks via Relevant Walks, IEEE Transactions on Pattern Analysis and Machine Intelligence <URL:https://arxiv.org/pdf/2006.03589.pdf>, Nov. 27, 2020, 20 pages. [cited by applicant]
Yuan et al., XGNN: Towards model-level explanations of graph neural networks, Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, <URL:https://dl.acm.org/doi/pdf/10.1145/339… [cited by applicant]
Khazane, Anish, et al. “Deeptrax: Embedding graphs of financial transactions.” 2019 18th IEEE International Conference on Machine Learning And Applications (ICMLA). <URL:arXiv:1907.07225v1.pdf> Jul. 16, 2019, 8 pages. [cited by applicant]
Van Belle, Rafaël, et al. “Inductive Graph Representation Learning for fraud detection.” Expert Systems with Applications <URL:https://creativecommons.org/licenses/by-nc-nd/4.0/>, 2022, 13 pages. [cited by applicant]
Huang et al. “Graphlime: Local Interpretable Model Explanations for Graph Neural Networks”, IEEE Transactions on Knowledge and Data Engineering. vol. 35, No. 7. Jul. 7, 2023, 5pages. [cited by applicant]
Hamilton et al. Inductive Representation Learning on Large Graphs. Advances in neural information processing systems 30. 31st Annual Conference on Neural Information Processing Systems <URL:https://arxiv.org/pdf/1706.02… [cited by applicant]
Leskovec, Dr. Jure. Large-scale Graph Representation Learning. 2017 IEEE International Conference on Big Data (BigData) Abstract, 1 page. [cited by applicant]
Jiang, Fei, et al. “Fi-grl: Fast inductive graph representation learning via projection-cost preservation.” 2018 IEEE International Conference on Data Mining (ICDM). IEEE, 2018, 6 pages. [cited by applicant]
Li, Yiming, et al. “Temporal Graph Representation Learning for Detecting Anomalies in E-payment Systems.” 2021 International Conference on Data Mining Workshops (ICDMW). IEEE, 2021, 8 pages. [cited by applicant]
Link prediction with Heterogeneous GraphSAGE (HinSAGE). <URL:https://stellargraph.readthedocs.io/en/latest/demos/link-prediction/hinsage-link-prediction.html>, 12 pages. [cited by applicant]
Node representation learning with Deep Graph Infomax <URL:https://stellargraph.readthedocs.io/en/latest/demos/embeddings/deep-graph-infomax-embeddings.html>, 10 pages. [cited by applicant]
Hamilton et al. Representation Learning on Graphs: Methods and Applications. <URL: https://arxiv.org/abs/1709.05584> Apr. 10, 2018, 24 pages. [cited by applicant]