The Performance of Support Vector Machine in Classifying Public Sentiment toward Student Suicide Cases

Abstract

The increasing popularity of social media has created a large volume of user-generated content that can be used to measure public opinion on sensitive social issues. A student suicide case that attracted much public attention was widely discussed on YouTube. The purpose of this study is to evaluate the performance of the Support Vector Machine (SVM) algorithm in classifying public sentiment towards the case by using comments collected from YouTube. For data collection, 5000 comments were scraped from a YouTube video uploaded to the Denny Sumargo channel using the YouTube Data API. The data collected were manually grouped into positive and negative sentiment categories. The sentiment analysis process included several steps of text pre-processing, such as cleansing, case normalization, tokenization, stop-word removal, and stemming. Feature extraction was done using Term Frequency-Inverse Document Frequency (TF-IDF), and the class imbalance was addressed using Synthetic Minority Over-sampling Technique (SMOTE). The dataset was split into a training and a testing set at 80:20. The experimental results show that the SVM model achieved 99.96% accuracy on the training set and 89.25% on the test set. In addition, the model produced balanced evaluation metrics, with accuracy, recall, and F1-score values close to 89%. The results indicate that the SVM algorithm is effective and robust for sentiment classification of Indonesian social media comments, especially on sensitive social issues. The study adds to the body of knowledge on machine learning–driven sentiment analysis to understand public responses in the digital sphere.