Deep Learning Approach for Bullying Classification on Twitter Social Media with Indonesian Language

Abstract

Cyberbullying is usually through social media intermediaries. The victim of cyberbullying will feel very depressed because of the wide spread of social media that can be seen and accessed by many people and also the privacy of the victim has no meaning, even all the shame and ugliness of the victim can be accessed by many people. The purpose of this study was to analyze the text documents on social media and then classify them into two classes, namely indications of bullying or cleanliness. Word2Vec and LSTM (Long Short Term Memory) will be combined in this classification model. Based on the testing phase, it can be concluded that there is still a lot of bullying on social media, especially on Twitter. This is evident from a large amount of Twitter data that 81.6% contains bullying words or sentences. The results of this study can be used as a basis for social media managers to take decisive action against bullies.