Evaluating Inter-Column Logical Relationships in Synthetic Tabular Data Generation

TL;DR

This paper introduces three new metrics to evaluate how well synthetic tabular data preserves logical relationships across columns, highlighting gaps in current generation methods.

Contribution

It proposes novel evaluation metrics for logical relationships in synthetic data and demonstrates their effectiveness on real-world datasets.

Findings

Existing methods often fail to maintain logical consistency and dependencies.

The new metrics reveal deficiencies in current data generation techniques.

Code for the evaluation metrics is publicly available.

Abstract

Current evaluations of synthetic tabular data mainly focus on how well joint distributions are modeled, often overlooking the assessment of their effectiveness in preserving realistic event sequences and coherent entity relationships across columns.This paper proposes three evaluation metrics designed to assess the preservation of logical relationships among columns in synthetic tabular data. We validate these metrics by assessing the performance of both classical and state-of-the-art generation methods on a real-world industrial dataset.Experimental results reveal that existing methods often fail to rigorously maintain logical consistency (e.g., hierarchical relationships in geography or organization) and dependencies (e.g., temporal sequences or mathematical relationships), which are crucial for preserving the fine-grained realism of real-world tabular data. Building on these insights, this study also discusses possible pathways to better capture logical relationships while modeling the distribution of synthetic tabular data. The code is available at https://github.com/Yunbo-max/TabLogicEval.