Identifying Network Influencers In Social Media: A Statistical Approach

Abstract

Over the last few decades, influence has been studied in both academic and professional fields, including marketing, political science, and sociology. It has become increasingly relevant in our society. Today, businesses care about influence patterns because influential people can convince a network of potential customers to buy their products. Campaign managers care because winning over influential sponsors or figures can help them gain support from voters, improving the chances that their political candidates will win in upcoming elections. However, studying influence patterns has been difficult because influence itself is a broad and subjective concept that is not easily quantified. It involves human choices and societal tendencies, so it cannot be understood from simple lab analysis or field experiments. But because of the importance surrounding influence, there have been plenty of theoretical findings on this topic. Furthermore, society has rapidly become more and more technologically oriented. In particular, social media has been prevalent, with over 2.5 billion users worldwide. Out of the various social networks, Twitter is a platform where information is publically communicated in real-time, often to a relatively wide audience. This study aims to calculate influence scores for Twitter users and compare the results to existing scoring systems. Another aim is to use machine learning algorithms to classify Twitter users, identifying them on their interests as well as their influence types (ex: student, professional, brand/corporation).