Ensemble Machine Learning for Stress Level Detection in Thesis Students

Thesis students represent a demographic prone to stress and anxiety. Thus, it is essential to recognize and resolve the issue promptly, as this may prevent more serious mental health complications. This study assessed the efficacy of ensemble machine learning techniques in measuring the stress levels of students engaged in thesis work. Data from 501 thesis students across seven universities were collected and examined. To determine the stress label, this study employed a modified version of the Student-Life Stress Inventory. As the baseline, five standard machine learning algorithms: Support Vector Machine, Random Forest, K-Nearest Neighbors, and Neural Network were utilized. This study implemented a pairwise correlation technique to assess the performance of the baseline algorithms in combination. The result demonstrated that the integrated ensemble of Random Forest and Naïve Bayes attained an outstanding accuracy of 0.9957, surpassing the individual baseline algorithms, which exhibited accuracy rates of $0.9829,0.9829$, 0.4957, 0.9701, and 0.9359 for Support Vector Machine, Random Forest, Naive Bayes, K-Nearest Neighbors, and Neural Network, respectively. It indicates that ensemble learning may identify stress in educational institutions, enabling timely intervention.