Exploratory Factor Analysis of Graphical Features for Link Prediction in Social Networks

Abstract

Social networks attract much attention due to their ability to replicate social interactions at scale. Link prediction, or the assessment of which unconnected nodes are likely to connect in the future, is an interesting but non-trivial research area. Three approaches exist to deal with the link-prediction problem: featurebased models, Bayesian probabilistic models, and probabilistic relational models. In feature-based methods, graphical features are extracted and used for classification. Usually, these features are subdivided into three feature groups based on their formula. Some formulas are extracted based on neighborhood graph traverse. Accordingly, there exist three groups of features: neighborhood features, pathbased features, and node-based features. In this paper, we attempt to validate the underlying structure of topological features used in feature-based link prediction. The results of our analysis indicate differing results from the prevailing grouping of these features, which indicates that current literatures' classification of feature groups should be redefined. Thus, the contribution of this work is exploring the factor loading of graphical features in link prediction in social networks. To the best of our knowledge, no prior studies had addressed it.

A novel similarity measure of link prediction in multi‐layer social networks based on reliable paths

“…We will now increase the link weight between users u 2 and u 5 to 0.8 to make this path more important for communication between these users in the future. Based on this, the similarity is calculated according to Equation (10).…”

“…Sim (u 2 , u 5 ) = [ 0.05 2 × ((0.4 × 0.9) + (0.3 × 0.8)) ] + [ 0.05 3 × (0.4 × 0.9 × 0.6) ] = 0.0015 (10) It is clear that the proposed similarity measure increases the likelihood of linking between users u 2 and u 5 in the future due to increased link weights. Therefore, this technique considers safe and strong communication between links in calculating similarity between users, which can be effective for link prediction.…”

“…Although, the link prediction problem has been extensively studied and various researches have been presented to solve it 7–9 ; However, the problem of how to optimally and effectively combine information to describe future communications remains largely unresolved. In Reference 10, to link prediction in social networks, the analysis of user's demographic features has been used. The results show that the cluster coefficient and the shortest path are effective in the link prediction.…”

Providing a new method for link prediction in social networks based on the meta-heuristic algorithm

“…Finally, the results of the comparison of the extracted features show that the clustering coefficient and the shortest path are effective in predicting the link with the accuracy criterion of 0.97. Aghabozorgi and Khayyambashi (2018), proposed a new similarity criterion for predicting graft based on local structures in social networks [24] . This similarity criterion is presented through a two-step peer learning model.…”

Altruistic and selfish communication on social media: the moderating effects of tie strength and interpersonal trust

“…Principal component analysis and other types of exploratory factor analysis are used for classifying intercorrelated variables under more general (latent) variables, something that is useful for reducing the dimensionality of data. For example, previous studies have used factor analysis to identify discrete dimensions of Facebook usage (Spiliotopoulos & Oakley, 2015) and network dimensions in social network graphs (Madahali, Najjar, & Hall, 2019). In the current study, in order to get an accurate composite measure of the text and photograph communication characterizing the friendships, a principal component analysis with orthogonal rotation (varimax)…”