Analisis Sentimen Cyberbullying Pada Komentar X Menggunakan Metode Naïve Bayes

Abstract

Advances in communication technology and social media have made it easier to access global information, but have also increased cases of cyberbullying on platforms such as X. The impact of cyberbullying can include physical and psychological disorders, such as increased loneliness, anxiety, depression, and decreased self-esteem. In addition, victims of cyberbullying may feel distress that can increase the risk of suicidal ideation. This research utilizes the Naïve Bayes method to effectively and efficiently classify cyberbullying-related comments. This classification model was developed to detect cyberbullying in comments on X, using the Naïve Bayes algorithm and a dataset from Kaggle consisting of 650 comments that contain cyberbullying characteristics and those that do not. This research includes several preprocessing steps such as tokenization, normalization, and stemming. The data was then divided into two parts: 80% for training data and 20% for testing data. The evaluation results show a model accuracy of 80.77%, precision 81.25%, recall 70.91%, and AUC 0.794. The innovation in this research lies in the use of 2 (two) stemming operators, namely stemming dictionary and stemming snowball, where the stemming dictionary uses a special file containing abbreviations or slang words, which are often used in comments on the word becomes its basic form. This model tends to be more accurate in classifying comments as non-bullying than bullying. Suggestions for improvement include exploring other preprocessing methods and algorithms, as well as using larger and more varied datasets.