Analisis Sentimen Terhadap Objek Wisata di Kabupaten Ngada Menggunakan Metode XGBoost

The natural beauty and rich culture of Ngada Regency in East Nusa Tenggara make it an attractive tourist destination, including Bena Traditional Village, Soa Mengeruda Hot Springs, and Manu Lalu Panorama. However, tourism management still faces various challenges, one of which is the lack of optimization of tourist attraction development. This research aims to understand tourist sentiment through reviews on Google Maps using the XGBoost method, a machine learning algorithm proven effective in text classification. Data was collected using web scraping technique using Octoparse application and successfully collected as many as 1,160 reviews for Bena Traditional Village tourism, 584 reviews for Soa Mengeruda Hot Spring, and 394 reviews for Manu Lalu Panorama. The data was then manually labeled by five reviewers. The data preprocessing stage includes cleaning, case folding, stopword removal, tokenization, and stemming, with text features converted using TF-IDF. The XGBoost model was optimized through Grid Search to find the best hyperparameter combination. The results showed that the XGBoost algorithm with optimal hyperparameters (n_estimators: 300, learning_rate: 0.3, min_child_weight: 1, max_depth: 3, subsample: 1.0) successfully classified sentiments with 84.6% accuracy, 83.9% precision, 84.6% recall, and 82.5% F1-score. The use of Grid Search was shown to significantly improve model performance compared to no hyperparameter optimization. The XGBoost method is effective in analyzing the sentiment of visitors' reviews of tourist attractions in Ngada Regency. The results of this analysis are expected to provide insight to managers and related parties in improving the quality of tourism management and development in the Ngada Regency area.