Systems and methods for data stream simulation

Patent metadata

Field	Value
Publication number	US-12455778-B2
Application number	US-201816151431-A
Country	US
Kind code	B2
Filing date	Oct 4, 2018
Priority date	Jul 6, 2018
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

A cloud computing system can be configured to generate a synthetic data stream that tracks a reference data stream. A model optimizer of the cloud computing system can receive, from an interface of the cloud computing system, a synthetic data stream request indicating a reference data stream. A dataset generator of the cloud computing system can generate a synthetic data stream that tracks the reference data stream by repeatedly swapping data models of the reference data stream. One such repeat can include retrieving, by the dataset generator from a model storage, a current data model of the reference data stream and generating a new data model of the reference data stream. The model optimizer can store the new data model in the model storage. The dataset generator can generate a synthetic data stream using the current data model of the reference data stream.

First claim

What is claimed is:

1. A cloud computing system for generating a synthetic data stream, comprising:
at least one processor; and
at least one non-transitory memory storing instructions that, when executed by the at least one processor, cause the cloud computing system to perform operations comprising:
receiving, from an interface, a request to generate a synthetic data stream, the request indicating a reference data stream from a stream source;
receiving, from the interface, a correlation metric for the reference synthetic data stream; and
generating a synthetic data stream that tracks the reference data stream, wherein the generating comprises repeatedly updating data models of the reference data stream at predetermined time intervals, a repeat comprising:
retrieving, from a model storage, a current data model of the reference data stream, the current data model comprising a generative adversarial network that generates an output resembling current characteristics of the reference data stream;
generating the synthetic data stream using the current data model of the reference data stream;
generating a new data model of the reference data stream;
evaluating performance criteria of the new data model by one or more duplicate elements in the synthetic data stream and the reference data stream, a prevalence of a common value in the synthetic data stream and the reference data stream, a maximum difference of rare values in the synthetic data stream and the reference data stream, and differences in schema between the synthetic data stream and the reference data stream;
storing, in the model storage, the new data model; and
updating, in the model storage, the current data model with the new data model.

2. He cloud computing system of claim 1 , wherein generating the new data model of the reference data stream comprises:
provisioning computing resources with the current data model; and
training the new data model on the computing resources using current reference data stream data.

3. The cloud computing system of claim 2 , wherein the repeat further comprises:
receiving reference data stream data; and
including the reference data stream data into the current reference data stream data upon receipt; or
storing the received reference data stream data, retrieving reference data stream data stored during a previous repeat, and including the retrieved reference data stream data into the current reference data stream data.

4. The cloud computing system of claim 1 , wherein the repeat occurs at one of a predetermined time or upon expiration of a time interval.

5. The cloud computing system of claim 1 , wherein the repeat occurs when a data schema of the reference data stream changes.

6. The cloud computing system of claim 1 , wherein the data models comprise recurrent neural networks and the reference data stream comprises JSON log data.

7. The cloud computing system of claim 6 , wherein generating the synthetic data stream using the current data model of the reference data stream comprises:
validating the synthetic data stream using a JSON validator and a schema for the reference data stream.

8. The cloud computing system of claim 7 , wherein: the schema describes key-value pairs present in the reference data stream; and
validating the synthetic data stream comprises validating that keys present in the synthetic data stream are present in the schema.

9. The cloud computing system of claim 7 , wherein:
the schema describes key-value pairs present in the reference data stream; and
validating the synthetic data stream comprises determining that key-value formats present in the synthetic data stream match corresponding key-value formats in the reference data stream.

10. The cloud computing system of claim 1 , wherein generating the synthetic data stream using the current data model of the reference data stream comprises:
identifying a sensitive portion of the reference data stream using a recurrent neural network;
generating a synthetic portion using the current data model; and
replacing the sensitive portion of the reference data stream with the synthetic portion.

11. The cloud computing system of claim 10 , wherein:
the current data model comprises a class-specific model corresponding to a data class;
identifying the sensitive portion of the reference data stream comprises determining that the sensitive portion of the reference data stream belongs to the data class; and
generating the synthetic portion comprises:
selecting the class-specific model based on the data class; and
generating the synthetic portion using the class-specific model.

12. The cloud computing system of claim 10 , wherein:
the current data model comprises a class-and subclass-specific model corresponding to a data class and a subclass of the data class; identifying the sensitive portion of the reference data stream comprises determining that the sensitive portion of the reference data stream belongs to the data class; and
generating the synthetic portion comprises: selecting the subclass; selecting the class- and subclass-specific model based on the data class and the selected subclass; and generating the synthetic portion using the class- and subclass-specific model.

13. The cloud computing system of claim 1 , wherein the repeat further comprises:
evaluating performance criteria of the new data model;
determining metadata of the new data model; and
storing the new data model and the metadata based on the evaluation of the performance criteria of the new data model.

14. The cloud computing system of claim 13 , wherein the performance criteria include at least one of a statistical correlation score, a data similarity score, a data quality score, a prediction accuracy check, a prediction accuracy cross check, a regression check, a regression cross check, or a principal component analysis check.

15. The cloud computing system of claim 13 , wherein the metadata includes at least one of an indication of an origin of the new data model, the origin corresponding to data used to generate the new data model, or when the new data model was generated.

16. The cloud computing system of claim 13 , wherein evaluating the performance criteria of the new data model comprises evaluating a number of matching elements in the synthetic data stream and reference data stream.

17. The cloud computing system of claim 1 , wherein the reference data stream includes sensitive portions of one or more datasets.

18. The cloud computing system of claim 13 , wherein evaluating the performance criteria of the new data model comprises comparing covariances or univariate distributions of a synthetic dataset generated by the new synthetic data model and a reference data stream dataset.

Assignees

• Capital One Services Llc

Inventors

• Watson Mark
• Truong Anh
• Abdi Taghi Abad Fardin
• Goodsitt Jeremy
• Walters Austin
• Walters Michael
• Tatsumi Noriaki
• Key Kate

Classifications

• G06T11/10

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06F18/213

  Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods · CPC title

• G06N3/094

  Adversarial learning · CPC title

• G06N3/06

  Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons · CPC title