Synthetic data generation apparatus, method for the same, and program

Patent metadata

Field	Value
Publication number	US-12265649-B2
Application number	US-201816753037-A
Country	US
Kind code	B2
Filing date	Oct 5, 2018
Priority date	Oct 13, 2017
Publication date	Apr 1, 2025
Grant date	Apr 1, 2025

Abstract

A synthetic data generation apparatus codes a value of each of category attributes contained in original data into a value of a numerical attribute in accordance with a coding rule; generates first synthetic data from the original data after coding using a synthetic data generation method for numerical attributes; if the value of the numerical attribute which is contained in the first synthetic data and corresponds to the value of one of the category attributes exceeds a range of values that can be assumed by the value of that numerical attribute, converts the value of that numerical attribute to a value included in the range of values that can be assumed by the value of that numerical attribute; and decodes the value of the numerical attribute which is contained in the first synthetic data after conversion and corresponds to the value of one of the category attributes to the value of that category attribute in accordance with the coding rule to obtain synthetic data.

First claim

What is claimed is:

1. A synthetic data generation apparatus comprising:
storage that stores a coding rule which indicates correspondence between a code and a value of a category attribute, and
processing circuitry configured to:
code a value of each of category attributes contained in original data into a value of a numerical attribute in accordance with the coding rule;
generate first synthetic data from the original data after coding using a synthetic data generation method for numerical attributes;
if the value of the numerical attribute which is contained in the first synthetic data and corresponds to the value of one of the category attributes exceeds a range of values that can be assumed by the value of that numerical attribute, convert the value of that numerical attribute to a value included in the range of values that can be assumed by the value of that numerical attribute; and
decode the value of the numerical attribute which is contained in the first synthetic data after conversion and which corresponds to the value of one of the category attributes to the value of that category attribute in accordance with the coding rule to obtain synthetic data, wherein
the value of the numerical attribute is a value that can be measured numerically, and the value of the category attribute is a value that cannot be measured numerically,
the coding rule is a 1-of-K coding method, and
the synthetic data maintains relationships among all the attributes in the original data, wherein
(i) the relationships are variance-covariance or (ii) the relationships are correlation.

2. A non-transitory computer-readable recording medium having recorded thereon a program for causing a computer to function as the synthetic data generation apparatus according to claim 1 .

3. A synthetic data generation method for execution by a synthetic data generation apparatus that includes storage and processing circuitry, the synthetic data generation method comprising:
a coding step in which the processing circuitry codes a value of each of category attributes contained in original data into a value of a numerical attribute in accordance with a coding rule which is stored in the storage and indicates correspondence between a code and a value of a category attribute;
a data formatting step in which the processing circuitry generates first synthetic data from the original data after coding using a synthetic data generation method for numerical attributes;
a conversion step in which, if the value of the numerical attribute which is contained in the first synthetic data and corresponds to the value of one of the category attributes exceeds a range of values that can be assumed by the value of that numerical attribute, the processing circuitry converts the value of that numerical attribute to a value included in the range of values that can be assumed by the value of that numerical attribute; and
a decoding step in which the processing circuitry decodes the value of the numerical attribute which is contained in the first synthetic data after conversion and which corresponds to the value of one of the category attributes to the value of that category attribute in accordance with the coding rule to obtain synthetic data, wherein
the value of the numerical attribute is a value that can be measured numerically, and the value of the category attribute is a value that cannot be measured numerically,
the coding rule is a 1-of-K coding method, and
the synthetic data maintains relationships among all the attributes in the original data, wherein
(i) the relationships are variance-covariance or (ii) the relationships are correlation.

Assignees

• Nippon Telegraph & Telephone

Inventors

• Okada Rina
• Hasegawa Satoshi
• Masaki Shogo
• Tanaka Satoshi

Classifications

• G06F16/258

  Data format conversion from or to a database · CPC title

• G06F16/285

  Clustering or classification · CPC title

• G06F21/6254 Primary

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