<![CDATA[The rapid growth of digital banking has increased user interaction through mobile banking apps such as BRImo (Bank Rakyat Indonesia). Google Play Store reviews provide valuable insight into app quality, but their unstructured format makes manual analysis inefficient. This study analyzes user sentiment toward BRImo and compares the performance of Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) for sentiment classification. Reviews were collected using Google Play Scraper from May 2024 to May 2025, yielding 15,945 raw reviews. After cleaning (removing duplicates, symbols, links, emojis) and language filtering, 15,233 valid reviews remained. Sentiment labels were generated using two lexicon-based methods: INSET and VADER. Using INSET, the data consisted of 6,238 positive, 4,987 negative, and 4,383 neutral reviews, producing 11,225 reviews for modeling. Using VADER, 10,496 positive, 2,903 negative, and 1,834 neutral reviews were obtained, totaling 13,399 reviews. Datasets were split into 80% training and 20% testing with stratified sampling. Features were extracted using TF-IDF unigrams. Classification was performed using linear SVM and KNN, with the optimal K=3 selected via Grid Search. Models were evaluated using 5-fold cross-validation, reporting mean accuracy, precision, recall, and F1-score (macro-average for INSET; weighted-average for VADER due to class imbalance). Results show SVM consistently outperforms KNN, achieving 98.36% mean accuracy and 98.34% mean F1-score on INSET, and 95.59% mean accuracy and 95.56% mean F1-score on VADER. Overall, BRImo user sentiment is predominantly positive, and findings can guide developers in improving app stability and quality]]>
