<![CDATA[Purpose: This study aims to analyze guest sentiment expressed in TripAdvisor reviews of Conrad Bali using the Multinomial Naïve Bayes algorithm to identify service attributes discussed by guests and generate evidence-based managerial insights that support hotel performance enhancement. Methodology: A quantitative text-mining approach was used. A total of 833 TripAdvisor reviews (2023–2025) were collected using the Apify web scraping tool.. The analysis involved text preprocessing, TF-IDF feature extraction, sentiment classification using the Multinomial Naïve Bayes algorithm, Synthetic Minority Oversampling Technique (SMOTE) for class balancing, confusion matrix evaluation, and word cloud visualization using Python. Results: The findings revealed that Positive, neutral, and negative sentiments accounted for 88.48%,8.64%and 2.88%, respectively. The application of SMOTE improved the model classification balance, increasing the F1-score from 0.8298 to 0.8596. Words related to staff, service, room, and hotel facilities appeared most frequently within positively classified reviews, indicating recurring discussion themes associated with favorable guest experiences. Conclusions: The findings demonstrate that sentiment analysis can effectively identify guest perceptions and recurring service attributes discussed in online hotel reviews. These insights support evidence-based service evaluation and managerial decision-making, thereby contributing to continuous service improvement and hotel performance enhancement. Limitations: The study was limited to TripAdvisor reviews, and the word clouds reflected word frequency rather than contextual meaning. In addition, the relatively small number of negative reviews may limit the classifier's ability to generalize minority sentiment patterns. Contributions: This study demonstrates how online guest reviews can be transformed into evidence-based managerial insights for service improvement, evaluation, and hotel performance enhancement.]]>
Copyrights © 2026
