Link Prediction Framework for Complex Networks Novel Similarity based Approaches

Abstract

In this thesis, we have presented five novel algorithms to solve the link prediction newlineproblem in different types of complex networks. The first algorithm uses the Peron newlineFrobenius theorem to compute the importance of node for link prediction in unipartite newlinegraphs. The second algorithm uses the concept of six degrees of separation and the total newlinenumber of possible walks between node pairs for the link prediction. The third algorithm newlineis for the unweighted and weighted networks, and this algorithm is based on the vertex newlineentropy and ego networks. The fourth algorithm is for unipartite, bipartite and weighted newlinenetworks that uses the concepts of Current-flow Centrality and the fifth algorithm is only newlinedesigned for the bipartite networks.