System and methods for generation of synthetic data cluster vectors and refinement of machine learning models

Patent metadata

Field	Value
Publication number	US-2021049456-A1
Application number	US-201916537884-A
Country	US
Kind code	A1
Filing date	Aug 12, 2019
Priority date	Aug 12, 2019
Publication date	Feb 18, 2021
Grant date	—

Abstract

Embodiments of the present invention provide an improvement to conventional machine model training techniques by providing an innovative system, method and computer program product for the generation of synthetic data using an iterative process that incorporates multiple machine learning models and neural network approaches. A collaborative system for receiving data and continuously analyzing the data to determine emerging patterns is provided. The proposed invention involves generating synthetic data clusters to be stored and used for retraining the main model as well as other models. In addition, the invention includes using one or more (subset) of the synthetic data clusters to train or retrain machine learning models, developing and training machine learning models that are trained with emerging synthetic data clusters, and ensembling machine learning models trained with emerging synthetic data clusters.

First claim

What is claimed is:

1 . A system for generation of synthetic data cluster vectors, the system comprising:
a module containing a memory storage device, a communication device, and a processor, with computer-readable program code stored thereon, wherein executing the computer-readable code is configured to cause the processor to:
receive input data for analysis and expansion;
analyze the input data using a machine learning model to identify an emerging pattern in the input data;
extract common data characteristics from the identified emerging pattern in order to determine a data scenario;
expand the data scenario via general adversarial neural network encoding to produce expanded data scenarios;
batch the expanded data scenarios to create multiple synthetic data sets containing varied subsets of the expanded data scenarios;
generate one or more syntenic data vectors comprising one or more of the multiple synthetic data sets; and
transmit the one or more synthetic data vectors to one or more machine learning models in order to train the models based on varied subsets of the expanded data scenarios.

2 . The system of claim 1 , wherein generating the multiple synthetic data sets further comprises refining the multiple synthetic data sets by:
processing the synthetic data set using the machine learning model to determine if the synthetic data set fits the identified emerging pattern; and iteratively narrowing scope of one or more of the multiple synthetic data sets based the results from the machine learning model.

3 . The system of claim 1 , wherein the system is further configured to:
determine one or more machine learning models as appropriate for the identified emerging pattern; and generate an ensemble of one or more machine learning models according to the identified emerging pattern.

4 . The system of claim 1 , wherein the machine learning model is retrained using a combination of real input data and one or more synthetic data vectors.

5 . The system of claim 1 , further configured to:
generate a machine learning model ensemble of the one or more machine learning models; and identify additional emerging patterns in the received input data.

6 . The system of claim 5 , wherein the machine learning model ensemble is continuously updated to include a subset of one or more additional machine learning models determined to meet a threshold of accuracy in identifying emerging patterns in received input data.

7 . The system of claim 1 , further comprising:
receiving a continuous input data feed; continually updating the emerging pattern detected by the machine learning model; extracting common data characteristics from the updated emerging pattern; determining if the data scenario requires adjustment based on the updated emerging pattern; and continually producing additional synthetic data vectors for the multiple synthetic data sets via the generative adversarial neural network.

8 . A computer-implemented method for iterative synthetic data generation, the computer-implemented method comprising:
receive input data for analysis and expansion; analyze the input data using a machine learning model to identify an emerging pattern in the input data; extract common data characteristics from the identified emerging pattern in order to determine a data scenario; expand the data scenario via general adversarial nerual network encoding to produce expanded data scenarios; batch the expanded data scenarios to create multiple synthetic data sets containing varied subsets of the expanded data scenarios; generate one or more syntenic data vectors comprising one or more of the multiple synthetic data sets; and transmit the one or more synthetic data vectors to one or more machine learning models in order to train the models based on varied subsets of the expanded data scenarios.

9 . The computer-implemented method of claim 8 , wherein generating the synthetic data set further comprises refining the multiple synthetic data sets by:
processing the multiple synthetic data sets using the machine learning model to determine if the synthetic data set fits the identified emerging pattern; and iteratively narrowing scope of one or more of the multiple synthetic data sets based the results from the machine learning model.

10 . The computer-implemented method of claim 8 , further comprising:
determining one or more machine learning models as appropriate for the identified emerging pattern; and generating an ensemble of one or more machine learning models according to the identified emerging pattern.

11 . The computer-implemented method of claim 8 , wherein the machine learning model is retrained using a combination of real input data and one or more synthetic data vectors.

12 . The computer-implemented method of claim 8 , further comprising:
generate a machine learning model ensemble of the one or more machine learning models; and identify additional emerging patterns in the received input data.

13 . The computer-implemented method of claim 12 , wherein the machine learning model ensemble is continuously updated to include a subset of one or more additional machine learning models determined to meet a threshold of accuracy in identifying emerging patterns in received input data.

14 . The computer-implemented method of claim 8 , further comprising:
receiving a continuous input data feed; continually updating the emerging pattern detected by the machine learning model; extracting common data characteristics from the updated emerging pattern; determining if the data scenario requires adjustment based on the updated emerging pattern; and continually producing additional synthetic data vectors via the generative adversarial neural network.

15 . A computer program product for iterative synthetic data generation, the computer program product comprising at least one non-transitory computer-readable medium having computer-readable program code portions embodied therein, the computer-readable program code portions comprising:
an executable portion configured for receiving input data for analysis and expansion; an executable portion configured for analyzing the input data using a machine learning model to identify an emerging pattern in the input data; an executable portion configured for extracting common data characteristics from the identified emerging pattern in order to determine a data scenario; an executable portion configured for expanding the data scenario via general adversarial neural network encoding to produce expanded data scenarios; an executable portion configured for batching the expanded data scenarios to create multiple synthetic data sets containing varied subsets of the expanded data scenarios; an executable portion configured for generating one or more syntenic data vectors comprising one or more of the multiple synthetic data sets; and an executable portion configured for transmitting the one or more synthetic data vectors to one or more machine learning models in order to train the models based on varied subsets of the expanded data scenarios.

16 . The computer program product of claim 15 , wherein generating the synthetic data set further comprises refining the synthetic data set by:
processing the multiple synthetic data sets using the machine learning model to determine if the synthetic data set fits the identified emerging pattern; and iteratively narrowing scope of one or more of the multiple synthetic data sets based the results from the machine learning model

17 . The computer program

Assignees

• Bank Of America

Inventors

• Kursun Eren

Classifications

• G06V10/82

  using neural networks · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06F18/23

  Clustering techniques · CPC title