Comparison of Naive Bayes and SVM Algorithms in Sentiment Analysis for the Optimization of Hotel Operational Services in Central Bangka Regency

<p><em>Customer reviews significantly influence hotel reputation and service improvement, but the large volume and unstructured nature make manual analysis inefficient. This study compares Naive Bayes and Support Vector Machine (SVM) for sentiment classification of 898 Indonesian-language hotel reviews from two online travel agencies, aiming to support service optimization in Central Bangka Regency. Reviews were preprocessed through cleaning, case folding, tokenization, stopword removal, and stemming, then converted into TF-IDF vectors. Both models performed well: SVM achieved the highest accuracy, precision, and F1-score, while Naive Bayes had the highest recall. This indicates that Naive Bayes is better at detecting more positive reviews, whereas SVM provides more precise and balanced classifications. The findings confirm the applicability of classical machine learning for hotel sentiment analysis and offer a basis for implementing real-time review monitoring systems in the hospitality sector.</em></p><p><strong><em>Keywords</em></strong><strong><em>:</em></strong><em> Sentiment Analysis; Naive Bayes; Support Vector Machine; Machine Learning; Hotel Customer Reviews</em></p><p align="center"><strong>Abstrak</strong></p><p>Ulasan pelanggan sangat memengaruhi reputasi hotel dan strategi peningkatan layanan, namun volume yang besar dan sifatnya yang tidak terstruktur membuat analisis manual menjadi tidak efisien. Penelitian ini membandingkan algoritma <em>Naive Bayes</em> dan <em>Support Vector Machine</em> (SVM) untuk klasifikasi sentimen dari 898 ulasan hotel berbahasa Indonesia yang dikumpulkan dari dua agen perjalanan daring, dengan tujuan mendukung optimasi layanan di Kabupaten Bangka Tengah. Ulasan diproses melalui tahap <em>cleaning, case folding, tokenization, stopword removal, </em>dan<em> stemming</em>, kemudian diubah menjadi vektor TF-IDF. Kedua model menunjukkan kinerja yang baik: SVM mencapai akurasi, presisi, dan F1-<em>score</em> tertinggi, sementara Naive Bayes memiliki recall tertinggi. Hasil ini menunjukkan bahwa Naive Bayes lebih baik dalam mendeteksi ulasan positif, sedangkan SVM memberikan klasifikasi yang lebih presisi dan seimbang. Temuan ini menegaskan penerapan algoritma machine learning klasik untuk analisis sentimen hotel dan menjadi dasar pengembangan sistem pemantauan ulasan secara real-time di sektor perhotelan.</p><p> </p>