<![CDATA[The transition of some Balinese Hindus away from the conventional ngaben cremation ceremony toward modern crematorium services has become a widely discussed public topic on social media platform X, largely centered on concerns about cost, time, and the potential erosion of customary values. This study labels netizen sentiment on this issue using a lexicon-based approach and builds and evaluates an automatic classification model based on Support Vector Machine (SVM), benchmarked against K Nearest Neighbor (KNN), as a preliminary step toward designing a public opinion monitoring information system. A total of 841 mixed language tweets were gathered from platform X and narrowed to 533 Indonesian language records after preprocessing (case folding and removal of URLs, mentions, hashtags, punctuation, and numbers). lexicon-based labelling, using a sentiment dictionary purpose built for the Balinese cultural context, produced a distribution of 50.66% neutral, 44.28% positive, and 5.07% negative sentiment. The data were then TF IDF weighted and split 80:20 into stratified training and testing sets. Testing showed the linear kernel SVM reaching 82.24% accuracy, far ahead of KNN (K=5) at only 52.34% on the same data. Per class evaluation showed SVM performing well on the neutral and positive classes but still limited on the negative class (33% recall) because of class imbalance. These results suggest SVM is more reliable than KNN for high dimensional, class imbalanced text data, and can serve as a foundation for an automatic classification module within a public opinion monitoring system supporting cultural preservation policy in Bali.]]>
Copyrights © 2026
