Abstract
Link prediction remains a challenging pursuit in existing complex networks. Our study proposes a Grover coin driven quantum walk (GCQW) model for prediction of missing edges on complex networks. The GCQW model uses observed probabilities of common neighbours of two nodes as similarity between the nodes. Furthermore, each walk step of the proposed model is determined by a three degree of influence rule. Results of experiments based on the area under the receiver operating characteristic curve (AUC) index demonstrate the proposed model's performance in eight real complex networks outperforms nine conventional comparison algorithms. Outcomes show that even when the ratio of testing to training is set in the range 0.1∼0.5, our GCQW model maintained a stable and competitive performance in terms of the AUC index. The proposed GCQW model will be expectedly applied in function modular mining of protein-protein interaction networks and friend recommendation of social media.