Exploring Innovative Approaches to Synthetic Tabular Data Generation

This comprehensive review explores data generation methodologies, focusing on statistical and machine learning-based approaches. Novel strategies like the divide-and-conquer (DC) approach and models such as GANBLR are examined, addressing challenges in preserving data relationships and enhancing interpretability. The integration of GANs is highlighted for its revolutionary impact across various sectors. The review analyzes how these techniques mitigate class imbalance, data scarcity, and privacy concerns, examining evaluation metrics and diverse applications. It concludes with insights into future research directions and the evolving role of synthetic data in advancing machine learning and data-driven solutions.

Eugenia Papadaki, Aristidis G. Vrahatis, Sotiris Kotsiantis