Prediksi Tingkat Stres Mahasiswa Selama Pembelajaran Daring Menggunakan Algoritma Machine Learning

Abstract

The sudden transition to online learning during the COVID-19 pandemic has had a significant psychological impact
on students, particularly in the form of increased stress levels. This study aims to
identify and analyze the factors that influence student stress during online learning
using a quantitative approach and predictive modeling. Data were obtained from 100 students aged 18–
25 years, covering variables such as screen time, sleep duration, physical activity, pre-exam anxiety, and changes in
academic performance. Statistical analysis showed that high screen time, less than 6 hours of sleep, and
academic anxiety were significantly associated with increased stress levels (p < 0.01). The Random Forest model
successfully predicted stress categories with 82% accuracy and identified sleep duration as the most
dominant factor. These findings indicate the need for more adaptive academic policy reforms regarding mental health,
including digital load management, healthy sleep education, and the integration of psychological support. This study
provides an empirical basis for educational institutions to design data-driven preventive interventions to
reduce the prevalence of stress among students.