Sentiment Analysis of Raya Digital Bank Application Reviews Using the TF-IDF Method and Support Vector Machine

This research examines sentiment analysis on user reviews of the Raya Digital Bank application, employing term frequency-inverse document frequency (TF-IDF) and support vector machine (SVM) techniques. As digital banking continues to rise in Indonesia, understanding customer feedback is critical for enhancing the user experience. The dataset, comprising user reviews scraped from the Google Play Store, underwent preprocessing steps, including cleansing, case folding, stopword removal, stemming, and tokenization. The TF-IDF method was applied to quantify word importance, converting text data into feature vectors, which were then classified using SVM. Four experimental scenarios were tested to optimize the model's performance: varying data splits (50:50 and 80:20), evaluating the impact of stemming, comparing Unigram and Bigram configurations, and testing different SVM kernels (linear, polynomial, RBF, and sigmoid). Results indicate that the best accuracy was achieved using a linear SVM kernel with an 80:20 data split, Unigram configuration, and without stemming, yielding a precision of 86.31%, recall of 85.37%, and F1-score of 85.67%. These findings indicate that combining Unigram-based TF-IDF and a linear SVM model effectively classifies sentiment in application reviews. The study recommends further tuning kernel parameters and testing trigrams to improve accuracy, especially for compound word usage common in app reviews. This approach provides a robust model for analyzing user sentiment, aiding digital banks in refining customer-oriented services.