<![CDATA[This study aims to analyze user review sentiments for six tourist attractions in East Nusa Tenggara Province by utilizing a large amount of review data obtained from Google Maps. Data was collected through a scraping process using Serp API, followed by cleaning and text pre-processing to improve data quality. Sentiment labeling was performed automatically using the Indo-BERT model to obtain three sentiment classes: positive, negative, and neutral. Text feature representation was performed using the Term Frequency–Inverse Document Frequency (TF-IDF) method, then classified using the baseline Support Vector Machine (SVM) model and the optimized SVM model with Grid-Search CV. The evaluation results showed that the baseline SVM model produced an accuracy of 83.87%, but showed an imbalance in performance between classes with a Macro F1-score of 0.4287. After parameter optimization using Grid-Search CV, the optimized SVM model produced an accuracy of 78.27% with an increase in the Macro F1-score value to 0.4818. This increase indicates an improvement in the model's ability to recognize minority sentiment classes despite a decrease in overall accuracy. Overall, the optimized SVM model provides more balanced and representative classification results in describing tourists' perceptions based on online reviews, so it can be used as a basis for sentiment analysis in the tourism sector.]]>
Copyrights © 2026
