Synthetic data generator evaluative visualization tool

Official URL: https://dx.doi.org/10.1109/SIU66497.2025.11111841

The increasing use of machine learning in sensitive domains has necessitated synthetic data generation to ensure privacy protection and data balance. This study introduces an innovative Synthetic Data Generator Evaluation Visualization Tool to address the complexities of creating and evaluating synthetic data. Our tool visualizes GAN model losses and performance metrics in real-time, enabling users to actively monitor and evaluate the training process. By progressively expanding GAN training with dynamic sampling of the dataset, our tool allows users to iteratively improve data convergence and efficiently adjust hyperparameters without using the entire dataset. The study validates the tool's effectiveness through rigorous testing and highlights potential improvements to extend its applicability. This research advances synthetic data studies by providing a scalable solution that simplifies the generation process, making a significant contribution to data-driven fields requiring synthetic datasets.