BiLSTM-Based Sentiment Analysis Of Traveloka Hotel Reviews In Yogyakarta For Data-Driven Communication Strategies

Online customer reviews have become a crucial medium of communication between guests and service providers in the hospitality industry. This study aims to perform sentiment analysis on hotel reviews from Traveloka to support data-driven customer communication strategies. Using a Bidirectional Long Short-Term Memory (BiLSTM) deep learning model, 10,681 user-generated reviews related to hotels in Yogyakarta were collected, preprocessed, and classified into binary sentiment categories. To address class imbalance, Synthetic Minority Oversampling Technique (SMOTE) and class weighting were applied. The model achieved 90.17% accuracy, 93.61% precision, 95.31% recall, and 94.45% F1-score, indicating strong generalization and sentiment recognition performance. The results highlight the model's ability to extract meaningful sentiment patterns, which can enhance hotel management’s responsiveness, improve communication strategies, and support continuous service improvement based on customer feedback.