Link prediction using matrix factorization with bagging

Link prediction aims at estimating the likelihood of the existence of links between nodes. In this paper, we treat link prediction as a collaborative filtering problem, and propose an algorithm to solve this problem using matrix factorization approach. For making better predictions, this paper also explores the use of bagging technique as combination approaches for matrix factorization. We subsample the training set and added random noise to make multiple classifiers, and then combine these classifiers to be the final classifier. Results on several data sets show the efficacy of our approach. Compared with several popular proximity metrics, the accuracy of the algorithm can be increased greatly by use of bagging technique, especially measured by precision.