# US 2021/0279564 A1 · Modifying data cleansing techniques for training and validating an artificial neural network model

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0279564 A1
- Document ID: US20210279564A1
- Status: Published application
- Kind: A1
- Application number: 16/809,198
- Filed: 2020-03-04
- Published: 2021-09-09
- Earliest claimed: 2020-03-04
- Assignees: HCL America Inc.
- Applicants: Hcl America, Inc.
- Inventors: Stevan ANTIC
- Web page: https://patentgrep.com/patent/US20210279564A1

## CPC

- G06N 3/08: Learning methods
- G06F 18/217: {Validation; Performance evaluation; Active pattern learning techniques}
- G06F 18/2413: based on distances to training or reference patterns
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Disclosed is a method and system for modifying data cleansing techniques for training and validating an Artificial Neural Network (ANN) model. The method comprises normalizing and validating data of different formats, obtained from different sources. The ANN model is trained using the normalized and validated data. Alternatively, the ANN model could be trained using data of a common format obtained from a deterministic data model. The trained ANN model is used to generate ANN stochastic data. Data validation component from the deterministic data model is reused for the normalizing and the validating of the data, for validating the ANN stochastic data.

## Claims

### Claim 1

A method for modifying data cleansing techniques for training and validating an Artificial Neural Network (ANN) model, the method comprising:
normalizing and validating data collected in different formats from different data sources;
training the ANN model using the normalized and validated data;
generating ANN stochastic data using the trained ANN model; and
reusing data validation component of a deterministic data model used for the normalizing and the validating of the data, for validating the ANN stochastic data.

### Claim 2 (depends on 1)

The method as claimed in claim 1, wherein the normalizing and the validating of the data includes:
normalizing the data present in different formats to a common format by mapping the different data sources to a common data source;
parsing the normalized data;
performing format validation and restriction validation on the parsed data; and
restarting parsing upon occurrence of an error till data of a required format is obtained.

### Claim 3 (depends on 1)

The method as claimed in claim 1, further comprising utilizing a deterministic data model for generating data of a common format.

### Claim 4 (depends on 3)

The method as claimed in claim 3, wherein OneTest Data is used as the deterministic data model to generate the data of a common format.

### Claim 5 (depends on 4)

The method as claimed in claim 4, wherein the common format is one of (Comma-Separated Values) CSV, JavaScript Object Notation (JSON), and Extensible Markup Language (XML).

### Claim 6 (depends on 1)

The method as claimed in claim 1, wherein the ANN is implemented using one of Recurrent Neural Network (RNN) classes comprising a Long Short-Term Memory (LSTM) and a Gated Recurrent Unit (GRU), Restricted Boltzmann Machines (RBM), and a Deep Belief Net (DBN).

### Claim 7

A system for modifying data cleansing techniques for training and validating an Artificial Neural Network (ANN) model, the system comprising:
a memory; and
a processor coupled to the memory, wherein the processor is capable of executing instructions to perform steps of:
normalizing and validating data collected in different formats from different data sources;
training the ANN model using the normalized and validated data;
generating ANN stochastic data using the trained ANN model; and
reusing data validation component of a deterministic data model used for the normalizing and the validating of the data, for validating the ANN stochastic data.

### Claim 8 (depends on 7)

The system as claimed in claim 7, wherein the normalizing and the validating of the data includes:
normalizing the data present in different formats to a common format by mapping the different data sources to a common data source;
parsing the normalized data;
performing format validation and restriction validation on the parsed data; and
restarting parsing upon occurrence of an error till data of a required format is obtained.

### Claim 9 (depends on 7)

The system as claimed in claim 7, further comprising utilizing a deterministic data model for generating data of a common format.

### Claim 10 (depends on 9)

The system as claimed in claim 9, wherein OneTest Data is used as the deterministic data model to generate the data of a common format.

### Claim 11 (depends on 10)

The system as claimed in claim 10, wherein the common format is one of (Comma-Separated Values) CSV, JavaScript Object Notation (JSON), and Extensible Markup Language (XML).

### Claim 12 (depends on 7)

The system as claimed in claim 7, wherein the ANN is implemented using one of Recurrent Neural Network (RNN) classes comprising a Long Short-Term Memory (LSTM) and a Gated Recurrent Unit (GRU), Restricted Boltzmann Machines (RBM), and a Deep Belief Net (DBN).

### Claim 13

A non-transitory computer program product having embodied thereon a computer program for modifying data cleansing techniques for training and validating an Artificial Neural Network (ANN) model, the computer program product storing instructions for:
normalizing and validating data collected in different formats from different data sources;
training the ANN model using the normalized and validated data;
generating ANN stochastic data using the trained ANN model; and
reusing data validation component from a deterministic data model used for the normalizing and the validating of the data, for validating the ANN stochastic data.