Systems and methods for synthetic data generation for time-series data using data segments

Patent metadata

Field	Value
Publication number	US-2020012902-A1
Application number	US-201916405989-A
Country	US
Kind code	A1
Filing date	May 7, 2019
Priority date	Jul 6, 2018
Publication date	Jan 9, 2020
Grant date	—

Abstract

Systems and methods for generating synthetic data are disclosed. For example, a system may include one or more memory units storing instructions and one or more processors configured to execute the instructions to perform operations. The operations may include receiving a dataset including time-series data. The operations may include generating a plurality of data segments based on the dataset, determining respective segment parameters of the data segments, and determining respective distribution measures of the data segments. The operations may include training a parameter model to generate synthetic segment parameters. Training the parameter model may be based on the segment parameters. The operations may include training a distribution model to generate synthetic data segments. Training the distribution model may be based on the distribution measures and the segment parameters. The operations may include generating a synthetic dataset using the parameter model and the distribution model and storing the synthetic dataset.

First claim

What is claimed is:

1 . A system for generating synthetic data, comprising:
one or more memory units storing instructions; and one or more processors that execute the instructions to perform operations comprising:
receiving a dataset comprising time-series data;
generating a plurality of data segments based on the dataset;
determining respective segment-parameters of the data segments;
determining respective distribution measures of the data segments;
training a parameter model to generate synthetic segment-parameters, the training being based on the segment parameters;
training a distribution model to generate synthetic data-segments, the training being based on the distribution measures and the segment parameters;
generating a synthetic dataset using the parameter model and the distribution model; and
storing the synthetic dataset.

2 . The system of claim 1 , wherein generating a synthetic dataset comprises:
generating, via the parameter model, a series of synthetic segment-parameters; and generating, via the distribution model, a series of synthetic data-segments based on the series of synthetic segment-parameters;

3 . The system of claim 1 , wherein:
the operations further comprise generating the parameter model; and training the parameter model is based on generating the parameter model.

4 . The system of claim 1 , wherein:
the operations further comprise generating the distribution model; and training the distribution model is based on generating the distribution model.

5 . The system of claim 1 , wherein generating data segments is based on a predetermined segment-size.

6 . The system of claim 1 , wherein generating data segments comprises determining a segment-size based on a statistical measure of the dataset.

7 . The system of claim 1 , wherein the segment parameters comprise a minimum value, a maximum value, a start value, and an end value.

8 . The system of claim 1 , wherein the distribution measures include at least one of a variance, a standard deviation, or a regression result of a time-dependent function.

9 . The system of claim 1 , wherein the parameter model comprises at least one of a recurrent neural network model or a long short-term memory model.

10 . The system of claim 1 , wherein the distribution model comprises a multilayer perceptron model, a convolutional neural network model, or a sequence-to-sequence model

11 . The system of claim 1 , wherein training the distribution model comprises:
training the distribution model to generate synthetic segment data; determining synthetic distribution-measures of the synthetic data-segments; determining a performance metric based on the distribution measures and the synthetic distribution-measures; and terminating training of the distribution model based on the performance metric satisfying a criterion.

12 . The system of claim 1 , wherein training the parameter model is based on an autocorrelation of the segment parameters.

13 . The system of claim 1 , the operations further comprising:
generating a data profile of the dataset; and storing the parameter model in a data index based on the data profile.

14 . The system of claim 1 , the operations further comprising:
generating a data profile of the dataset; and storing the distribution model in a data index based on the data profile.

15 . The system of claim 1 , wherein:
the dataset comprises multidimensional time-series data; the data segments comprise multidimensional data segments; and the segment parameters comprise multidimensional segment parameters.

16 . The system of claim 15 , wherein:
the operations further comprise determining a correlation of two or more dimensions of the segment-parameters; and training the parameter model is based on the correlation.

17 . The system of claim 1 , wherein:
receiving the dataset comprises receiving the dataset from a client device; and the operations further comprise transmitting the synthetic dataset to the client device.

18 . The system of claim 1 , wherein receiving the dataset comprises receiving the dataset at a cloud service.

19 . A method for generating synthetic data, the method comprising:
receiving a dataset comprising time-series data; generating a plurality of data segments based on the dataset; determining respective segment parameters of the data segments; determining respective distribution measures of the data segments; training a parameter model to generate synthetic segment-parameters, the training being based on the segment parameters; training a distribution model to generate synthetic data-segments, the training being based on the distribution measures and the segment parameters; generating a synthetic dataset using the parameter model and the distribution model; and storing the synthetic dataset.

20 . A system for generating synthetic data, comprising:
one or more memory units storing instructions; and one or more processors that execute the instructions to perform operations comprising:
receiving a dataset comprising time-series data;
generating a data profile of the dataset;
generating a plurality of data segments based on the dataset;
determining respective segment-parameters of the data segments, the segment parameters comprising a minimum value, a maximum value, a start value, and an end value;
determining respective distribution measures of the data segments;
generating a parameter model based on the dataset;
training the parameter model to generate synthetic segment-parameters, the training being based on the segment parameters;
generating a distribution model based on the dataset;
training the distribution model to generate synthetic data-segments, the training being based on the distribution measures and the segment parameters;
generating a synthetic dataset by:
generating, via the parameter model, a series of synthetic segment-parameters; and
generating, via the distribution model, a series of synthetic data-segments based on the series of synthetic segment-parameters;

storing the synthetic dataset; and
storing the parameter model and the distribution model in a data index based on the data profile.

Assignees

• Capital One Services Llc

Inventors

• Walters Austin
• Watson Mark
• Truong Anh
• Goodsitt Jeremy
• Farivar Reza
• Key Kate
• Pham Vincent
• Rafferty Galen

Classifications

• G06F9/541 Primary

  via adapters, e.g. between incompatible applications · CPC title

• G06F21/6254

  by anonymising data, e.g. decorrelating personal data from the owner's identification · CPC title

• G06T11/10

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

• G06F18/214

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

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title