<![CDATA[The expansion of online travel agencies (OTAs) has produced large volumes of user-generated hotel reviews, offering important resources for sentiment analysis of consumer perceptions. However, prior studies largely rely on single-platform datasets and focus on classification performance, with limited attention to cross-platform sentiment consistency and the impact of data imbalance. This study aims to analyse and compare sentiment patterns across Traveloka, Tiket.com, and Accor, while evaluating a machine learning framework under imbalanced data conditions. This study adopts a quantitative experimental design using 3,000 Indonesian-language reviews collected via web scraping. The independent variable is reviewing text, and the dependent variable is sentiment classification (positive/negative). Data were preprocessed and transformed using TF-IDF, and classified using Multinomial Naïve Bayes, with performance evaluated by accuracy, precision, recall, and F1-score. The results show that positive sentiment consistently dominates across all platforms, with Accor achieving the highest performance, followed by Tiket.com and Traveloka. However, very high recall values for the positive class indicate substantial class imbalance, which biases predictions and reduces sensitivity to negative sentiment. This study provides empirical evidence of cross-platform sentiment consistency and highlights the importance of addressing data imbalance in sentiment modelling.]]>
Copyrights © 2026
