<![CDATA[Floods are one of the most common types of natural disasters in the Sumatra region and have sparked a variety of public reactions that are widely shared on social media, particularly on the X (Twitter) platform. This study aims to analyze public perceptions of the flood disaster in Sumatra using the Support Vector Machine (SVM) algorithm. Data was collected by gathering tweets containing the hashtags #prayforsumatera, #PrayForSumatera, #prayforsumatra, and #PrayforSumatra from November 27, 2025, to January 13, 2026, totaling 1,364 tweets in Indonesian. The research process involved data preprocessing (cleaning, lowercasing, tokenization, removal of irrelevant words, and stemming), automatic sentiment analysis using the VADER lexicon and standard thresholds (composite score ≥ 0.05 = positive; ≤ -0.05 = negative; in-between = neutral), feature extraction based on TF-IDF (3,000 features, n-grams 1–2), data splitting for training and testing with an 80:20 ratio, and classification modeling via an SVM optimized with GridSearchCV. The labeling results show a sentiment distribution of 603 positive data points (44.21%), 499 negative data points (36.58%), and 262 neutral data points (19.21%). The best parameters were found with C=1, class weight=‘balanced’, and a linear kernel, yielding a CV F1-macro score of 0.6260. The classification model developed achieved an accuracy of 71% on 273 test data points, with an average F1-score of 0.68 and a weighted average of 0.71. These findings indicate that SVM can identify public sentiment regarding the flood disaster in Sumatra with reasonably good performance, although categorization into the neutral class is more difficult due to the ambiguous nature of the text.]]>
Copyrights © 2026
