Revealing the political affinity of online entities through their Twitter followers

What is it about?

In this work, we show that the structural features of the Twitter online social network can divulge valuable information about the political affinity of the participating nodes. More precisely, we show that Twitter followers can be used to predict the political affinity of prominent Nodes of Interest (NOIs) they opt to follow. We utilize a series of purely structure-based algorithmic approaches, such as modularity clustering, the minimum linear arrangement (MinLA) problem and the DeGroot opinion update model in order to reveal diverse aspects of the NOIs’ political profile. Our methods are applied to a dataset containing the Twitter accounts of the members of the Greek Parliament as well as an enriched dataset that additionally contains popular news sources. The results confirm the viability of our approach and provide evidence that the political affinity of NOIs can be determined with high accuracy via the Twitter follower network. Moreover, the outcome of an independently performed expert study about the offline political scene confirms the effectiveness of our methods.

Featured Image

Photo by Merakist on Unsplash

Photo by Merakist on Unsplash

Why is it important?

Our approaches are simple to implement and easy to reproduce while delivering very significant accuracy. Furthermore, the overlap coefficient, an underutilized measure, especially in the context of social networks, is highlighted and shown to achieve a considerably superior efficiency compared to other projection functions. Additionally, we applied concepts to this problem that have not been examined in prior literature, the MinLA problem and the DeGroot model with the presence of stubborn nodes, and showed that these techniques are perfectly suitable for the analysis of our dataset. We deem that the application of these novel ideas will have a theoretical impact on future research regarding Social Network Analysis. Our methods work without any prior knowledge of the ground truth related to the NOIs and do not employ heavy preprocessing or filtering on the raw data.