Data generation apparatus and method thereof

A data generation apparatus comprising:

• a data input unit configured to input a first dataset;
• a data preprocessing unit configured to normalize the first dataset, and split the normalized initial dataset into a initial training dataset and a initial test dataset;
• a learning model generation unit configured to generate a first learning model by training a first machine learning model using the initial training dataset and optimizing the first machine learning;
• a validation unit configured to validate the generated first learning model using the first test dataset;
• a semi-synthetic data generation unit configured to select and generate a new parameter within a boundary space defined by the initial dataset, and perform data prediction using the first learning model based on the new parameter to generate a semi-synthetized dataset;
• a semi-synthetic data filter unit configured to determine a fitness of the new parameter based on a second dataset generated by combining the semi-synthetized dataset and the first dataset and the first machine learning model, and determine whether the semi-synthetized dataset is to be accepted as valid data or discarded according to the fitness;
• and
• a database configured to store the initial dataset and the semi-synthetized dataset.

A data generation method performed by a data generation apparatus, the data generation method comprising:

• inputting a initial dataset;
• normalizing the initial dataset and splitting the normalized initial dataset into a initial training dataset and a initial test dataset;
• training a first machine learning model using the initial training dataset and optimizing a hyperparameter of the first machine learning model using a predetermined number of cross-validations, thereby generating a first learning model;
• validating the first learning model using the first test dataset;
• selecting and generating a new parameter from within a boundary space defined by the initial dataset;
• and
• performing data prediction based on the new parameter using the first learning model to generate a semi-synthetized dataset.