BUZZER DETECTION ON TWITTER USING MODIFIED EIGENVECTOR CENTRALITY

• Main Author: Tressa Juzar - NIM : 23516086 , Mario
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28760
• Institution: Institut Teknologi Bandung

Social media is an online media where its users can easily participate, share, and create content including blogs, social networks, wikis, forums and virtual worlds. One of the biggest social media is twitter. On twitter and other social media, there is a phenomenon which called buzzer. Buzzer is a social media user that has influence into other users. This influence can be seen from the number of interactions and responses from tweet by the buzzer. From friendship graph on social media, buzzer is considered as the central point of information dissemination or the central poin of friendship on social media. Based on this, to detect buzzer on twitter can be approached by doping centrality analysis. Centrality analysis has several calculation methods, namely degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality. However, in previous studies the calculation of the central point was limited to the static graph of the user without including the dynamics that occur on social media. The dynamics of social media, for example, how the interaction on a tweet and how other users respond to a tweet. In this thesis, a calculation method based on eigenvector centrality was developed by involving dynamics on social media which was then named modified eigenvector centrality. Modified eigenvector centrality is then used for the calculation of buzzer values. In addition, a simple software was also developed to detect buzzer using the developed calculations. Testing is conducted on software and buzzer value calculation. The software test is conducted by testing the functionality and the correctness of the calculation process. And for the buzzer value calculation test are conducted by testing the effects of variables on calculation, complexity and execution time of calculation, and test of spreading information based on recommended buzzer from calculation.