<![CDATA[Anxiety is a psychological condition frequently experienced by vocational high school students due to academic pressure, practical training demands, and uncertainty about future careers. This study aims to (1) classify the anxiety levels of vocational students based on their personal narratives shared via WhatsApp conversations, and (2) compare the performance of two kernel types in the Multi-class Support Vector Machine (SVM) classification model. This quantitative study used a computational experimental design involving 670 Grade X students from a vocational school in Gianyar, Bali. A total of 1,476 narrative texts were collected and labeled into five anxiety levels based on the DASS-42 scale: normal, mild, moderate, severe, and very severe. The classification process applied TF-IDF vectorization and compared the Radial Basis Function (RBF) and Sigmoid kernels. Evaluation results showed that the Sigmoid kernel achieved the highest accuracy (81.42%) and macro-average F1-score (0.7914), demonstrating better performance in recognizing minority classes. The model successfully identified students with severe (10.5%) and very severe (9.1%) anxiety, supporting its potential use for early psychological screening. These findings confirm that Multi-class SVM is effective for classifying anxiety levels from digital narratives and can be integrated into school-based mental health monitoring systems.]]>
Copyrights © 2025
