Analisis Performa Support Vector Machine (SVM) dengan Hyperparameter Tuning untuk Klasifikasi Data Stress Monitoring Siswa

Abstract

Stress is one of the critical issues in student mental health that directly affects academic performance. Machine learning–based stress detection offers an effective solution for automatically monitoring psychological conditions. This study analyzes the performance of the Support Vector Machine (SVM) algorithm in classifying student stress levels using a public dataset. Two approaches are compared: a baseline SVM model with a linear kernel and a model refined through hyperparameter tuning using GridSearchCV. The evaluation employed accuracy, precision, recall, F1-score, confusion matrix, and the McNemar statistical test. The results show that both models achieved identical performance, with an accuracy of 84.54% and a macro F1-score of 0.85. The confusion matrix further demonstrated identical prediction distributions, although the McNemar test indicated a significant difference (p < 0.05). These findings highlight that hyperparameter tuning does not necessarily lead to performance improvements, particularly when the data is nearly linearly separable. The implications of this research emphasize the importance of deeper analysis in selecting tuning strategies, while also opening opportunities to apply SVM models as the backend of mobile-based student stress monitoring applications.