Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Computer Science

Supervisor

Sedig, Kamran

Abstract

Twitter is one of the most popular microblogging and social networking services. Many people from a wide range of backgrounds use Twitter to contribute their thoughts on different topics through postings, known as ``tweets”. Analysts collect and analyze tweets to extract knowledge. To rely on tweets, it is crucial to assess Twitter users’ credibility. In recent years, researchers have proposed various techniques, especially data analytics models, for evaluating Twitter users and analyzing their behaviour; however, these techniques do not engage analysts in the process, leading to a lack of understanding and trust in results. In this thesis, an exploratory visual analytics system is designed and implemented to help with triaging Twitter users. To this end, the system can leverage analysts’ expertise and knowledge through interactive visualization to assist them in understanding the underlying information within data. Subsequently, a case study demonstrates the capabilities of the system in identifying Twitter users.

Summary for Lay Audience

As one of the most popular microblogging and social media services, Twitter has millions of monthly active users who publish an abundance of daily postings. This platform allows users from a wide range of backgrounds to publish their thought and opinion on various topics. Analyzing Twitter users' behaviour in terms of contributions to different topics and their group association is an integral part of investigating tweets. For instance, many automatic accounts, known as bots, are being used to propagate different content on Twitter. In addition, some associated accounts may publish fake news on Twitter. Therefore, to rely on tweets as a source of information, a journalist who aims to collect tweets for news and stay updated has to analyze and identify the users who posted about specific topics. The current approaches for identifying Twitter users do not engage the analysts in the process, leading to a lack of understanding and trust in results. This thesis proposes a system that represents information through interactive visualizations. Besides, the system takes advantage of data analysis models to support users' identification process. A case study shows that the system assists analysts in understanding Twitter users' information and detect the hidden patterns and potential similarities between the users.

Download

Share

COinS