Enhancing Understanding of Public Sentiment on Twitter Using SVM and Lexicon Methods
Dewi Suci Khairani, Andi Setiawan, Siti Masruroh · 2024
Indonesia has experienced 655 terrorism incidents between 2000 and 2024, leading to numerous casualties. This study aims to enhance understanding of public sentiment on Twitter in response to notable terrorism events, specifically the suicide bombing of the Astana Anyar Police Station in Bandung in 2022. Using a Support Vector Machine (SVM) algorithm for classification and two lexicon-based methods (VADER and InSet) for sentiment analysis, the study compared their performance and accuracy. The VADER lexicon indicated predominantly negative sentiments, while the InSet lexicon showed more positive sentiments, likely due to differences in dictionary sizes, data preparation, and feature extraction methods. The model using the VADER-labeled dataset achieved a 94% accuracy rate, compared to 93% for the InSet-labeled model. These high accuracy levels demonstrate the models proficiency in categorizing sentiment data, providing valuable insights into public reactions to significant events and underscoring the importance of selecting appropriate lexicon-based methods for sentiment analysis in social media contexts.