<![CDATA[While sentiment analysis is increasingly applied in hospitality research, existing studies predominantly rely on large, balanced datasets from global platforms or computationally intensive deep learning models that lack interpretability for local hotel management. A critical research gap remains in deploying lightweight, interpretable machine learning frameworks on small, highly imbalanced Indonesian hotel reviews while translating sentiment outputs into actionable operational insights. To address this gap, this study evaluates customer satisfaction at Hotel Puri Ansel using an XGBoost-based sentiment classification pipeline optimized for real-world data constraints. Google Reviews were processed through comprehensive Indonesian text preprocessing and TF-IDF feature extraction, then partitioned into 80% training and 20% testing sets. The XGBoost model achieved 84% accuracy and a 0.83 weighted F1-score, demonstrating exceptional positive sentiment recall (97%). Lexical analysis identified “cleanliness” and “comfort” as primary satisfaction drivers, whereas “hygiene issues” and “slow service” dominated negative feedback. Although the model exhibited limitations with ambiguous and sarcastic expressions, its novelty lies in bridging technical classification performance with interpretable business intelligence. This study contributes a reproducible, resource-efficient framework that enables local hospitality operators to leverage unstructured review data for targeted service improvements, prioritizing practical deployment validity over artificial data balancing.Keywords: Sentiment Analysis, XGBoost, Hotel Service Quality]]>
Copyrights © 2026
