Abstract:ABSTRACT Background Link prediction is an important and well-studied problem in network biology. Recently, graph representation learning methods, including Graph Convolutional Network (GCN)-based node embedding have drawn increasing attention in link prediction. Motivation An important component of GCN-based network embedding is the convolution matrix, which is used to propagate features across the network. Existing algorithms use the degree-normalized adjacency matrix for this purpose, as this matrix is closely related to the graph Laplacian, capturing the spectral properties of the network. In parallel, it has been shown that GCNs with a single layer can generate more robust embeddings by reducing the number of parameters. Laplacian-based convolution is not well suited to single-layered GCNs, as it limits the propagation of information to immediate neighbors of a node. Results Capitalizing on the rich literature on unsupervised link prediction, we propose using node similarity-based convolution matrices in GCNs to compute node embeddings for link prediction. We consider eight representative node-similarity measures (Common Neighbors, Jaccard Index, Adamic-Adar, Resource Allocation, Hub- Depressed Index, Hub-Promoted Index, Sorenson Index and Salton Index) for this purpose. We systematically compare the performance of the resulting algorithms against GCNs that use the degree-normalized adjacency matrix for convolution, as well as other link prediction algorithms. In our experiments, we use three-link prediction tasks involving biomedical networks: drug–disease association prediction, drug–drug interaction prediction and protein–protein interaction prediction. Our results show that node similarity-based convolution matrices significantly improve the link prediction performance of GCN-based embeddings. Conclusion As sophisticated machine-learning frameworks are increasingly employed in biological applications, historically well-established methods can be useful in making a head-start. Availability and implementation Our method, SiGraC, is implemented as a Python library and is freely available at https://github.com/mustafaCoskunAgu/SiGraC.

CrimeGraphNet: Link Prediction in Criminal Networks with Graph Convolutional Networks

CrimeGNN: Harnessing the Power of Graph Neural Networks for Community Detection in Criminal Networks

DeepOffense: a Recurrent Network Based Approach for Crime Prediction

CrimeGAT: Leveraging Graph Attention Networks for Enhanced Predictive Policing in Criminal Networks

TransCrimeNet: A Transformer-Based Model for Text-Based Crime Prediction in Criminal Networks

Deep Learning Criminal Networks

Xgcn: an Extreme Graph Convolutional Network for Large-scale Social Link Prediction

A Directed Link Prediction Method Using Graph Convolutional Network Based on Social Ranking Theory

HAGEN: Homophily-Aware Graph Convolutional Recurrent Network for Crime Forecasting

Addressing Crime Situation Forecasting Task with Temporal Graph Convolutional Neural Network Approach

Network Link Connectivity Prediction Based on GCN and Differentiable Pooling Model

HDM-GNN: A Heterogeneous Dynamic Multi-view Graph Neural Network for Crime Prediction

Identifying Linked Fraudulent Activities Using GraphConvolution Network

TransGCN:Coupling Transformation Assumptions with Graph Convolutional Networks for Link Prediction

Dynamic Causal Graph Convolutional Network for Traffic Prediction

Community Preserving Adaptive Graph Convolutional Networks for Link Prediction in Attributed Networks

Dual Convolutional Neural Network for Graph of Graphs Link Prediction

Complex Graph Convolutional Network for Link Prediction in Knowledge Graphs

Node similarity-based graph convolution for link prediction in biological networks

ComGCN: Community-Driven Graph Convolutional Network for Link Prediction in Dynamic Networks

A Survey on Graph Classification and Link Prediction based on GNN