Graph-Based Feature Augmentation for Predictive Tasks on Relational Datasets

2025-08-28

AI Summary

Existing feature enhancement methods for relational data prediction rely heavily on manual intervention, limiting scalability and automation. Method: This paper proposes ReCoGNN, an end-to-end automated framework that semantically partitions multi-table relational data into coherent fragments, constructs a heterogeneous weighted graph to model cross-table and cross-row dependencies, and jointly optimizes information propagation and feature selection via coupled graph neural networks. It integrates attribute-relation modeling, heterogeneous graph construction, and differentiable message passing to automatically extract and fuse high-order semantic features. Contribution/Results: Extensive experiments across 10 real-world and synthetic datasets demonstrate that ReCoGNN significantly outperforms state-of-the-art feature enhancement approaches on both classification and regression tasks, exhibiting strong generalization capability. The framework establishes a novel paradigm for automated feature engineering in relational data analytics.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Reasoning under Uncertainty: Relational Probabilistic Models
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Data has become a foundational asset driving innovation across domains such as finance, healthcare, and e-commerce. In these areas, predictive modeling over relational tables is commonly employed, with increasing emphasis on reducing manual effort through automated machine learning (AutoML) techniques. This raises an interesting question: can feature augmentation itself be automated and identify and utilize task-related relational signals? To address this challenge, we propose an end-to-end automated feature augmentation framework, ReCoGNN, which enhances initial datasets using features extracted from multiple relational tables to support predictive tasks. ReCoGNN first captures semantic dependencies within each table by modeling intra-table attribute relationships, enabling it to partition tables into structured, semantically coherent segments. It then constructs a heterogeneous weighted graph that represents inter-row relationships across all segments. Finally, ReCoGNN leverages message-passing graph neural networks to propagate information through the graph, guiding feature selection and augmenting the original dataset. Extensive experiments conducted on ten real-life and synthetic datasets demonstrate that ReCoGNN consistently outperforms existing methods on both classification and regression tasks.

Problem

• Automating feature augmentation for relational datasets
• Capturing semantic dependencies across multiple tables
• Enhancing predictive modeling through graph neural networks

Innovation

• Automated feature augmentation using relational tables
• Graph neural networks for information propagation
• Heterogeneous weighted graph representing inter-row relationships