Empirical Analysis of Traditional Link Prediction Methods

Online Social Networks are growing exponentially due to which a lot of researchers are working on Social Network analysis. Link Prediction is a task of predicting new links that may occur in future in the social network. The link prediction problem has generated a lot of interest due its widespread applicability across many do-mains. We conducted a study on the different methods that have been developed for link prediction. In most of these methods, the social network is modeled as a graph, and the links are predicted based on the similarities between two nodes. We have chosen seven widely used similarity methods in our study. We found that on the simulated data sets, Sorenson index method and Jaccard co-efficient method performed well when compared to other methods