IMPLEMENTASI METODE BERT DAN SVM PADA ANALISIS SENTIMEN GAME GENSHIN IMPACT

The Genshin Impact game has become a global phenomenon with a large player base, especially in Indonesia, which is the 4th largest user country in the world. This study aims to analyze user sentiment towards the game by utilizing data from the social media platform X. The analysis was carried out by comparing two sentiment classification methods, namely Support Vector Machine (SVM) and Bidirectional Encoder Representations from Transformers (BERT). Data was collected through a crawling process using API X and processed through preprocessing stages, such as cleansing, tokenization, and stemming. The SVM method was chosen because of its simplicity in implementation, while the BERT method was used to explore the ability of deep learning to understand complex linguistic contexts. This study shows that BERT provides higher classification accuracy than SVM, especially in handling the diversity of language styles on social media. It is hoped that the results of this research can provide input for game developers to improve user experience through events that are more in line with community preferences.