Synthetic Data Generation for Enhanced Model Efficiency: A GAN- Based Approach to Tabular Data

The generation of synthetic data has emerged as a promising approach to enhance the performance of machine learning models, particularly in scenarios with limited original data. This study focuses on using Conditional Tabular Generative Adversarial Networks (CTGANs) for synthetic tabular data generation. The synthetic tabular data produced is supposed to have the same statistical properties as the original data. The aim of this research is to examine the effect of augmented datasets that combine original and synthetic data on the performance of several machine learning models, namely Logistic Regression, Random Forest, and Gradient Boosting Machines. Experimental results reveal how model accuracy, precision, recall, and F1-score vary with the inclusion of synthetic data, thus giving insights into its effectiveness in improving model generalization. Statistical tests, like paired t-tests, validate the significance of performance differences, and experiments varying the proportion of synthetic data provide guidelines for its optimal usage. This work highlights the potential and limitations of synthetic data augmentation in practical machine learning applications. Besides, this paper is based on previous work that applied CTGANs to mitigate data imbalance in machine learning.