Generating Synthetic Relational Tabular Data via Structural Causal Models

2025-07-04

AI Summary

Existing synthetic tabular data methods struggle to model cross-table causal dependencies prevalent in real-world relational databases, leading to distorted and practically limited synthetic data. This paper proposes the first structural causal model (SCM)-based framework for relational tabular data generation, explicitly representing inter-table dependencies as a causal graph and integrating probabilistic graphical models with conditional generative mechanisms to enable joint multi-table modeling. It is the first approach to achieve interpretable and controllable synthesis of complex causal structures extracted from real relational databases. Experiments demonstrate that the generated data significantly outperforms state-of-the-art baselines in structural fidelity, statistical consistency, and downstream task performance—including TabPFN training and evaluation—thereby substantially enhancing the realism and practical utility of synthetic relational data.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Statistical Relational/Logic Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Synthetic tabular data generation has received increasing attention in recent years, particularly with the emergence of foundation models for tabular data. The breakthrough success of TabPFN (Hollmann et al.,2025), which leverages vast quantities of synthetic tabular datasets derived from structural causal models (SCMs), demonstrates the critical role synthetic data plays in developing powerful tabular foundation models. However, most real-world tabular data exists in relational formats spanning multiple interconnected tables - a structure not adequately addressed by current generation methods. In this work, we extend the SCM-based approach by developing a novel framework that generates realistic synthetic relational tabular data including causal relationships across tables. Our experiments confirm that this framework is able to construct relational datasets with complex inter-table dependencies mimicking real-world scenarios.

Problem

• Generating synthetic relational tabular data
• Addressing complex inter-table dependencies
• Extending SCMs for multi-table causality

Innovation

• Extends SCMs for relational tabular data
• Generates inter-table causal relationships
• Mimics real-world complex dependencies