A Data-Driven Machine Learning for Predicting Student Stress Levels

Abstrak

Academic stress significantly impacts students' psychological well-being and academic performance. This study focuses on predicting students' stress levels using a data-driven machine learning framework. The dataset was obtained from a questionnaire comprising 25 indicators encompassing emotional, psychological, academic, and environmental aspects of students. The research procedure involved data preprocessing, checking for missing values and redundancy, normalization, descriptive statistical analysis, model development, and performance evaluation using metrics such as recall, precision, sensitivity, specificity, F-measure, and accuracy. The implemented algorithm achieved excellent results, with an overall accuracy of 0.98. The model demonstrated high effectiveness in classifying Eustress and Distress, while its performance in detecting the No Stress category was limited, although precision and specificity indicate a strong capacity to differentiate between classes. These findings confirm that a machine learning approach can effectively capture patterns of student stress based on questionnaire responses and offers valuable guidance for developing early warning systems and targeted psychological intervention strategies. The study highlights the potential of data-driven predictive methods in supporting students' mental health through empirical data analysis.

Keywords - LibSVM; Machine Learning; Predicting; Stress Levels; Tree Ensemble.