Ranking based comparative analysis of graph centrality measures to detect negative nodes in online social networks

Online Social Networks (OSNs) are emerging as a communication platform where interaction among users results in the formation of positive or negative relations. Due to the existence of negative relations, many nodes are suspicious of masquerading and conspiring against popular nodes as well as intruding into the private groups of networks. Many researchers have analyzed negative nodes in networks of positive and negative ties by using measures such as degree, status, PII and PN centrality. While the existing literature focused only on small offline datasets, in this work an approach to identify negative nodes in large datasets of OSNs is proposed. The deviation of results of measures from actual behavior is examined using statistical and graphical techniques. It was observed that PN centrality measure is able to detect a number of outsiders of the network with higher accuracy as compared to other measures. However, some crucial nodes which are actually outsiders are misclassified as the most popular nodes by it. To counter this drawback, we have proposed and compared new values of parameters of PN measure for large-scale networks through graphs as well as statistical measures.