Directed Link Prediction using GNN with Local and Global Feature Fusion

TL;DR

This paper introduces a novel GNN framework that fuses feature embeddings with community information and transforms graphs into directed line graphs, significantly improving directed link prediction performance.

Contribution

It proposes a hybrid feature fusion method and a graph transformation technique to enhance directed link prediction with theoretical and empirical validation.

Findings

Outperforms state-of-the-art methods on benchmark datasets

Effective with varying amounts of training data (30%-60%)

Demonstrates the benefit of hybrid feature integration in GNNs

Abstract

Link prediction is a classical problem in graph analysis with many practical applications. For directed graphs, recently developed deep learning approaches typically analyze node similarities through contrastive learning and aggregate neighborhood information through graph convolutions. In this work, we propose a novel graph neural network (GNN) framework to fuse feature embedding with community information. We theoretically demonstrate that such hybrid features can improve the performance of directed link prediction. To utilize such features efficiently, we also propose an approach to transform input graphs into directed line graphs so that nodes in the transformed graph can aggregate more information during graph convolutions. Experiments on benchmark datasets show that our approach outperforms the state-of-the-art in most cases when 30%, 40%, 50%, and 60% of the connected links are used as training data, respectively.