A Link Prediction Algorithm Combining Node Influence and New Biased Random Walk

Recent years have seen a lot of interest in link prediction. Essentially, it means designing a prediction algorithm that is capable of accurately describing a certain network mechanism in order to get more accurate predictions. In complex network research, it has important applications. The DeepWalk method randomly samples neighbor nodes, and it does not fully consider the node position itself, so insufficient information considered in node sequence sequence sampling When using DeepWalk for link prediction, the node's own influence characteristics are not considered. First of all,we change DeepWalk through biased random walk, learns the vector representation of nodes, and then fuses the influence between node pairs and the similarity between node pairs to raise the accuracy of link prediction. We have experimented the algorithm of this paper on real network data, and from the experimental results, we can see that the algorithm of this paper works better than other link prediction algorithms.