Link Prediction for Complex Networks via Random Forest

In this paper, based on four existing similarity indexes (CN, LHN-II, COS+ and MFI), we obtain eigenvectors by extracting the features of two arbitrary complex network nodes. The core idea of this paper is to use decision trees to handle these four different indexes which are not strongly related. After training and learning with the random forest algorithm, a new link prediction algorithm for complex networks is proposed. We prove by conducting some numerical simulations, using the US aviation network as an example, that the proposed link prediction algorithm is more accurate and stable than other similar algorithms.