A Novel Centrality-based Measure for Election Network Analysis

Abstract

With the worldwide adaptation of the internet, social network platforms have been rapidly engaging a vast global population. Such networks are subject to user engagement and interaction on various topics - social, political, economic, etc. The present work gathers such interaction between Twitter users over a period of 6 weeks, where the topic of interaction was restricted to the West Bengal state assembly election. The conversion of the acquired data to a node adjacency list was performed on the basis of the mention relationship between users. The network analysis was performed using a novel measure based on only the eigen-vector centrality of the users of the network, termed the NetworkPresenceFactor(NPF). The results reveal that network users affiliated with Party_1 had the most influence and presence over the dynamic network, followed by those affiliated with Party_2 and Party_3. The analysis of party-wise network influence also reveals a similar scenario. The weekly trends of presence reveal that Party_1 had a constantly reducing presence that was almost twice that of Party_2, while Party_3 started off with a minimum influence that peaked during week 6. On tallying, the results of our experiment differ from the ground truth. The authors conclude that the consideration of network structure alone (as in our work) is not sufficient in effective analysis of the network and the inclusion of the context of user interaction is essential for proper network analysis.