Advances in digital technology allow the use of social media as a source of public opinion, including in the tourism sector. This study analyzed visitor sentiment towards Matayangu Waterfall in Central Sumba with a combined lexicon-based approach and Naive Bayes algorithm. Comment data was taken from TikTok through the web crawling method for the period January 1, 2024 to April 30, 2025. Data is processed through text preprocessing stages such as cleaning, normalization, and stemming. The initial sentiment label was determined using the SentiWord-ID dictionary, and then converted to numerical forms using TF-IDF before being classified by Naive Bayes. Evaluations are carried out to measure the performance of the model. The results of the evaluation showed that the Naïve Bayes method had an accuracy of 75.81% in predicting sentiment. Most of the comments exhibited a neutral sentiment (61.69%), with positive comments accounting for 20.67% and negative ones for 17.63%.
Nggiri et al. (Wed,) studied this question.