<![CDATA[The development of digital technology in Indonesia has driven the growth of application-based transportation services such as Gojek. However, amid this growth, user dissatisfaction with pricing strategies often appears in social media reviews, but has not been systematically explored as a basis for optimizing delivery pricing using an automated text mining approach. Nevertheless, empirical studies that explicitly utilize sentiment analysis as a foundation for delivery pricing optimization remain limited. Existing sentiment analysis studies on online transportation services have mostly focused on overall service satisfaction, with limited attention to how sentiment patterns reflect customer price sensitivity. This study aims to analyze user sentiment as a diagnostic basis for evaluating and supporting delivery pricing optimization strategies. A comparative evaluation of four text representation methods: TF-IDF, Bag-of-Words (BoW), Word2Vec, and TF-IDF with Chi-Squared feature selection was conducted using 282 Gojek-related reviews collated from X (Twitter). The results of the experiment show that the Naïve Bayes model with BoW representation achieved the best performance, with 70% accuracy and an F1-score of 0.68, outperforming more complex approaches. Further analysis identifies that negative sentiment correlates with the keywords “expensive” and “promo,” providing a strong indication that pricing issues are a major point of concern for customers. These findings serve as a basis for management to optimize dynamic pricing strategies based on real-time user feedback.]]>
