Heterogeneous Line Graph Neural Network for Link Prediction.

Heterogeneous network link prediction is an important network information mining problem. Existing link prediction methods for heterogeneous networks typically require predefined meta-paths with prior knowledge. To address the problem, we propose a new model, named Heterogeneous Line Graph Neural Network (HLGNN), in this paper. Firstly, we design a line graph transformation module to encapsulate node features and transform the heterogeneous network into a heterogeneous line graph. Then, we propose an intra-type aggregation component to collect the same type of edges. As we have aggregated node information in each type, we design an inter-layer aggregation to combine messages from multiple node types. Finally, we put the aggregation results into a multilayer perceptron to achieve link prediction. The experimental results show that, compared to the state-of-the-art baselines, the proposed method achieves superior performance.