Penerapan Algoritma Support Vector Machine (SVM) Untuk Mengukur Tingkat Stres Mahasiswa Semester Akhir

Abstract

Abstract

This study focuses on the development of an academic stress classification system for university students undergoing the thesis-writing process using the Support Vector Machine (SVM) algorithm and comparing its performance with Random Forest. The dataset consisted of responses from 150 final-year students collected through a Likert-scale questionnaire covering four dimensions of academic stress: physiological, cognitive, behavioral, and emotional aspects. Data preprocessing, validity testing, and reliability testing were conducted before the modeling stage. The research methodology applied SVM with the Radial Basis Function (RBF) kernel and Random Forest as a comparative algorithm. Hyperparameter optimization was performed using Grid Search, while model validation employed Stratified K-Fold Cross Validation (k=10). Model performance was evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The results showed that SVM outperformed Random Forest, achieving an accuracy of 96.00%, precision of 96.31%, recall of 95.07%, F1-score of 94.56%, and ROC-AUC of 1.000, while Random Forest achieved an accuracy of 94.00%. The distribution of stress levels indicated that 66% of respondents experienced moderate stress. Furthermore, the emotional aspect was identified as the most influential factor contributing to academic stress intensity. These findings demonstrate that SVM is effective for academic stress classification and has potential to support the development of early stress detection systems in higher education institutions.

Keywords:

Academic Stress; Classification; Final-Year Students; Random Forest; Support Vector Machine

Abstrak

Pengembangan sistem klasifikasi stres akademik bagi mahasiswa yang sedang menempuh proses penyusunan skripsi menjadi fokus utama penelitian ini, dengan menempatkan algoritma Support Vector Machine (SVM) sebagai metode utama yang dikomparasikan dengan Random Forest. Basis data penelitian bersumber dari 150 mahasiswa akhir yang memberikan respons terhadap kuesioner berskala Likert, mencakup indikator-indikator dari empat domain,  fisiologis, kognitif, perilaku dan emosional. Integritas data dijamin melalui prosedur pra-pemrosesan yang sistematis, diikuti verifikasi validitas dan reliabilitas instrumen sebelum tahap pemodelan dilakukan. Kerangka metodologis penelitian ini mengintegrasikan SVM dengan fungsi kernel Radial Basis Function (RBF) dan Random Forest sebagai algoritma pembanding, didukung oleh teknik Grid Search untuk optimasi hiperparameter serta Stratified K-Fold Cross Validation (k=10) sebagai prosedur validasi model. Performa kedua algoritma dievaluasi secara komprehensif menggunakan metrik accuracy, precision, recall, F1-score, dan ROC-AUC. Temuan empiris mengindikasikan keunggulan SVM atas Random Forest, dengan capaian accuracy sebesar 96,00%, precision 96,31%, recall 95,07%, F1-score 94,56%, dan ROC-AUC bernilai sempurna 1,000, sementara Random Forest mencatatkan accuracy 94,00%. Distribusi tingkat stres responden memperlihatkan proporsi terbesar, yakni 66% pada kategori stres sedang. Analisis terhadap kontribusi dimensi mengungkapkan bahwa aspek emosional menjadi faktor paling determinan dalam membentuk intensitas stres akademik mahasiswa. Berdasarkan keseluruhan temuan tersebut, algoritma SVM terbukti efektif sebagai instrumen identifikasi stres akademik dan membuka peluang pengembangannya sebagai fondasi sistem deteksi dini stres di lingkungan perguruan tinggi.

Kata kunci:

Klasifikasi; Mahasiswa akhir; Random Forest; Stres Akademik; Support Vector Machine