Rel-MOSS: Towards Imbalanced Relational Deep Learning on Relational Databases

TL;DR

This paper introduces Rel-MOSS, a novel GNN-based approach to address class imbalance in relational databases by using relation-wise gating and relation-guided minority over-sampling, significantly improving classification performance.

Contribution

The paper pioneers the investigation of class imbalance in relational databases and proposes Rel-MOSS, combining relation-centric gating and minority synthesis for better entity classification.

Findings

Rel-MOSS outperforms SOTA methods with up to 2.46% and 4.00% improvements.

Extensive experiments on 12 datasets validate the effectiveness of the approach.

Rel-MOSS effectively mitigates minority class under-representation in RDBs.

Abstract

In recent advances, to enable a fully data-driven learning paradigm on relational databases (RDB), relational deep learning (RDL) is proposed to structure the RDB as a heterogeneous entity graph and adopt the graph neural network (GNN) as the predictive model. However, existing RDL methods neglect the imbalance problem of relational data in RDBs and risk under-representing the minority entities, leading to an unusable model in practice. In this work, we investigate, for the first time, class imbalance problem in RDB entity classification and design the relation-centric minority synthetic over-sampling GNN (Rel-MOSS), in order to fill a critical void in the current literature. Specifically, to mitigate the issue of minority-related information being submerged by majority counterparts, we design the relation-wise gating controller to modulate neighborhood messages from each individual relation type. Based on the relational-gated representations, we further propose the relation-guided minority synthesizer for over-sampling, which integrates the entity relational signatures to maintain relational consistency. Extensive experiments on 12 entity classification datasets provide compelling evidence for the superiority of Rel-MOSS, yielding an average improvement of up to 2.46% and 4.00% in terms of Balanced Accuracy and G-Mean, compared with SOTA RDL methods and classic methods for handling class imbalance.