I In the era of technological advancement, the internet plays a pivotal role as the primary source of global information, particularly in searching for details about tourist destinations. Tripadvisor, as a leading platform, enables users to provide and explore reviews and comments from fellow travelers, including those about destinations like Waterbom Bali. This research aims to analyze visitor sentiment through the implementation of the Random Forest and Naïve Bayes methods. Data was collected through the scrapping of Waterbom Bali reviews, yielding a total of 5750 entries. Despite data imbalances, class imbalance techniques were employed. Sentiment analysis was conducted by comparing the effectiveness of the Random Forest and Naïve Bayes methods using the Word2Vec feature extraction technique. Experimental results revealed significant differences between the two methods. For Random Forest, after undersampling, the accuracy reached 24%, while oversampling improved accuracy to 98%. Regarding Multinomial Naïve Bayes, after undersampling, accuracy reached 36%, and oversampling resulted in an accuracy of 97%.