ANALISIS SENTIMEN ULASAN APLIKASI MYBCA MENGGUNAKAN PELABELAN INDOBERT DAN KLASIFIKASI SUPPORT VECTOR MACHINE

User reviews on the MyBCA application available on Google Play Store contain various opinions that can be utilized to understand users' perceptions of the application. However, the continuously increasing number of reviews makes manual analysis inefficient, thereby requiring an automated sentiment analysis approach. This study aims to implement IndoBERT as an automatic sentiment labeling method and Support Vector Machine (SVM) as a sentiment classification algorithm for MyBCA user reviews on Google Play Store. The dataset consisted of 5,000 Indonesian-language reviews collected through web scraping. The research stages included data preprocessing, namely case folding, cleaning text, and normalization text, followed by sentiment labeling using a pre-trained IndoBERT model that classified reviews into positive, neutral, and negative sentiments. The dataset was then divided into 80% training data and 20% testing data. Feature extraction was performed using TF-IDF and n-gram techniques to transform text into numerical representations before classification using SVM. The results showed that negative sentiment dominated the reviews with 57.7%, followed by positive sentiment with 33.3%, and neutral sentiment with 9.0%. The evaluation results demonstrated that the SVM model achieved an accuracy of 90.4%, precision of 89.63%, recall of 90.4%, and F1-score of 89.91%. These findings indicate that the combination of IndoBERT, TF-IDF, and SVM is capable of effectively classifying sentiment in MyBCA user reviews, particularly for positive and negative sentiment classes.