Structured Evaluation of Synthetic Tabular Data

TL;DR

This paper introduces a unified evaluation framework for synthetic tabular data based on the principle that generated data should match the distribution of real data, enabling better assessment of data quality and generator performance.

Contribution

It proposes a coherent, mathematically grounded evaluation framework that unifies existing metrics and introduces new model-free baselines for synthetic tabular data quality assessment.

Findings

Structured synthesizers outperform deep learning methods on small datasets

The framework reveals the completeness and limitations of existing metrics

Model-based and model-free metrics are unified under a single objective

Abstract

Tabular data is common yet typically incomplete, small in volume, and access-restricted due to privacy concerns. Synthetic data generation offers potential solutions. Many metrics exist for evaluating the quality of synthetic tabular data; however, we lack an objective, coherent interpretation of the many metrics. To address this issue, we propose an evaluation framework with a single, mathematical objective that posits that the synthetic data should be drawn from the same distribution as the observed data. Through various structural decomposition of the objective, this framework allows us to reason for the first time the completeness of any set of metrics, as well as unifies existing metrics, including those that stem from fidelity considerations, downstream application, and model-based approaches. Moreover, the framework motivates model-free baselines and a new spectrum of metrics. We evaluate structurally informed synthesizers and synthesizers powered by deep learning. Using our structured framework, we show that synthetic data generators that explicitly represent tabular structure outperform other methods, especially on smaller datasets.