ANALISIS SENTIMEN TERHADAP CYBERBULLYING PADA TWITTER MENGGUNAKAN METODE GATED RECURRENT UNIT

The widespread use of social media, especially Twitter, has led to an increase in cases of Cyberbullying which attacks various personal aspects of users. This research aims to conduct sentiment analysis on cases of Cyberbullying on Twitter by grouping data into 5 target classes, namely ethnicity, age, gender, religion, and not bullying. The method applied in this research is Gated Recurrent Unit (GRU), which is effective in managing data sequences and
capturing long-term relationships in text. The research process included data collection via Twitter, data preprocessing (such as cleaning, tokenization, and stopword removal), as well as
applying word embedding techniques to convert text into a numerical representation.
The test results show that the GRU model used was successful in classifying tweets with an accuracy of 92%, proving that this method is effective in detecting and identifying types of Cyberbullying. It is hoped that the results of this research can become a reference in developing an early detection system to minimize the spread of negative content on social media and create
a safer and more comfortable digital environment.