Evaluating Inter-Column Logical Relationships in Synthetic Tabular Data Generation
This work addresses a gap in synthetic data evaluation for data scientists and practitioners, but it is incremental as it builds on existing methods by adding new metrics without introducing a novel generation approach.
The paper tackled the problem of evaluating synthetic tabular data generation by focusing on overlooked logical relationships across columns, proposing three new metrics and showing that existing methods often fail to maintain consistency, with experimental results revealing failures in preserving hierarchical and temporal dependencies crucial for realism.
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.