Emotion Detection on Fraudulent Accounts Using BERT and Kleinberg’s Burst Algorithms in Indonesian Text

The rapidly evolving information and communication technology aids people in their daily tasks, but it has also negative consequences, particularly the misuse of technology for criminal activities. There have been various methods to overcome these crimes, yet it often like a game of hide and seek. Cyber-enabled fraudsters exploit social media to manipulate victims into adopting their viewpoints without clear explanations. This paper analyzes Twitter posts of fraudsters to assess their influence on victims and compares them to those of general users. The analysis employs SLSA and the BERT model to determine sentiment polarity, as well as Kleinberg’s burst algorithm to identify sentiment surges. The results indicate that fraudsters exhibit a stronger inclination towards specific sentiments, whereas ordinary accounts show a more balanced sentiment distribution.