<![CDATA[This research analyzed user sentiment toward the NTB Syariah application using Support Vector Machine (SVM) and Naïve Bayes classification methods. A dataset comprising 814 reviews was obtained via web scraping, with 245 allocated for testing. Preprocessing encompassed cleaning, case folding, tokenization, filtering, and stemming, while sentiment labeling employed a lexicon-based approach integrated with TF-IDF weighting, categorizing reviews as positive, neutral, or negative. Model performance was assessed through accuracy, precision, recall, and F1-score metrics. Results demonstrated SVM's superior performance (accuracy: 92.65%; precision: 0.9327; recall: 0.9265; F1-score: 0.9149) compared to Naïve Bayes (accuracy: 84.49%; precision: 0.8415; recall: 0.8449; F1-score: 0.8005). SVM exhibited greater robustness in managing high-dimensional, complex, and moderately imbalanced datasets, delivering consistent cross-class sentiment classification. Conversely, Naïve Bayes remained computationally efficient and suitable for rapid implementation scenarios. These findings underscore machine learning's efficacy in sentiment analysis for digital banking platforms.]]>
Copyrights © 2025
