Implementasi Model Fine-Tuning BERT dalam Deteksi Konten Judi Online di Media Sosial Twitter (Aplikasi X)

Abstract

The rapid growth of social media usage has had a significant impact on society, particularly in terms of social interaction, information dissemination, and economic activity. However, behind these benefits, social media has also become a platform for the spread of harmful and unlawful content, including online gambling. Many users are influenced by such content, especially on the X Application platform. Therefore, this study aims to develop a web-based system capable of detecting online gambling-related content in tweets using the IndoBERT language model through a fine-tuning approach. The data used in this study were collected from Indonesian-language tweets related to online gambling activity on the X Application (Twitter). A total of 2,498 tweets were collected and divided into 70% training data (1,954 tweets) and 30% testing data (544 tweets). The evaluation results show that the model can identify online gambling and non-gambling content with an accuracy of 91%, precision of 94.5%, recall of 97.1%, and an F1-score of 95.3%.