Link Prediction in Directed Graphs via Dummy Node Line Graphs

Link prediction plays a critical role in graph mining fields such as social networks, bioinformatics, and recommendation systems. However, existing methods predominantly focus on undirected graphs, with insufficient attention given to directed graphs, which are more common in practical applications. This paper presents a novel Graph Neural Network-based approach for link prediction in directed graphs, incorporating a transformation of line graphs with dummy nodes combined with a Graph Attention Network. This method preserves essential graph structure information through the use of dummy nodes, effectively overcoming the issue of information loss associated with traditional line graph transformations. Furthermore, we propose a fast dummy node line graph transformation algorithm that reduces computational complexity, making it suitable for large-scale graph data applications. Our study provides new insights into directed graph link prediction and holds significant implications for practical applications, including recommendation systems, path optimization, and information dissemination.