Virtual Community Detection Through the Association between Prime Nodes in Online Social Networks and Its Application to Ranking Algorithms

Existing studies have contributed immensely to link prediction by identifying different types of network communities. In this paper, a new type of network community in online social networks (OSNs) is identified using the association between network nodes. This new network community is called "virtual community." Virtual communities are based on either the real/ physical relationships of users that are connected to their constituency, social, and professional activities or their virtual interactions associated with their cognitive levels, choice selection, and ideology. Users belonging to the same virtual community exhibit similar behavior in linking to nodes of common interest. These nodes, which reflect the common interest of a community, are called "prime nodes." Prime nodes are linked to the prediction problem in OSN completion and are generally recommended for OSN growth. Recent studies on ranking algorithms have shown that the incompleteness of OSNs contributes to the low accuracy of ranking algorithms in identifying top spreaders. Thus, in this paper, we propose an OSN completion method based on link prediction through association between prime nodes. An experiment on predicting new links in two real big data sets of two global OSNs, namely, Facebook and Twitter, is conducted. The effectiveness of the proposed method is also validated by applying prominent ranking algorithms to the newly predicted and original networks. Results show that the accuracy rates of the ranking algorithms are improved, thereby validating the importance of the proposed method in predicting vital links.