Spectral clustering for link prediction in social networks with positive and negative links

Abstract

Online social networks (OSNs) recommend new friends to registered users based on local features of the graph (i.e. based on the number of common friends that two users share). Real OSNs (e.g. Facebook) do not exploit all network structure. Instead, they consider only pathways of maximum length 2 between a user and his candidate friends. This can limit the accuracy of prediction. On the other hand, there are global approaches, which detect the overall path structure in a network, being computationally prohibitive for huge-size social networks. In this paper, we provide friend recommendations, by performing multi-way spectral clustering, which uses information obtained from the top few eigenvectors and eigenvalues of the normalized Laplacian matrix and computes a multi-way partition of the data. As a result, it produces a less noisy matrix, which is smaller and more compact than the original one, focusing on main linking trends of the social network. Thus, we are able to provide fast and more accurate friend recommendations. Moreover, spectral clustering compared to traditional clustering algorithms, such as k-means and DBSCAN, which assume globular (convex) regions in Euclidean space, is more flexible, in capturing the non-connected components of a social graph and a wider range of cluster geometries. We perform an extensive experimental comparison of the proposed method against existing link prediction algorithms, the k-means and two-way nCut clustering algorithms, using synthetic and three real data sets (Hi5, Facebook and Epinions). Our experimental results show that our SpectralLink algorithm outperforms the local approaches, the kmeans and two-way nCut clustering algorithms in terms of effectiveness, whereas it is more efficient than the global approaches. We show that a significant accuracy improvement can be gained by using information about both positive and negative edges.

Trust in Social Media

“…• Spectral clustering based on Laplacian matrix [54] for trust/distrust networks is performed first [124] and then define two similarities:…”

Predicting interval time for reciprocal link creation using survival analysis

“…Over the years researchers have solved the link prediction problem for a variety of graphs -for example link prediction in homogeneous networks (Hasan et al, 2006;Liaghat et al, 2013;Wang et al, 2017b), link prediction in heterogeneous information networks (Sun et al, 2011;Dong et al, 2012), and link prediction for knowledge graphs (Dong et al, 2014;Zhang et al, 2016). Other related problems, such as link/sign prediction and ranking in signed social network (Song and Meyer, 2015;Symeonidis and Mantas, 2013), and a recommendation system using link prediction techniques (Esslimani et al, 2011) have also been studied.…”

Exploiting Emotional Information for Trust/Distrust Prediction

“…Therefore, [20] is a special case of the algorithm in [6] when k = 3. Unsupervised methods are usually based on certain topological properties of trust and distrust networks to perform predictions [16,36,29]. One type of unsupervised methods is based on lowrank matrix factorization [16,36].…”

“…[36] extends the low-rank model to perform link prediction across multiple signed networks. Node similarity based trust and distrust prediction is another type of unsupervised methods [29,30], which first define similarity metrics to calculate node similarities; and then provide a way to predict trust and distrust relations based on those node similarities. Propagation-based methods are also used for trust and distrust prediction [9,14,37].…”