Strezzlah: An AI-Powered Stress Classification System for University Students using Machine Learning

Abstract

University students face increasing mental health challenges, with limited access to professional counseling services. This paper presents Strezzlah, an AI-powered web application that classifies student stress levels using machine learning algorithms. The system utilizes the standardized DASS-21 (Depression, Anxiety, and Stress Scale) questionnaire combined with Random Forest and XGBoost classifiers to provide real-time stress assessment and personalized recommendations. Implemented using Flask framework, the system achieved 87.41% accuracy with weighted F1 score of 87.02%. The platform serves students, counselors, and administrators through role-based interfaces, enabling early intervention and scalable mental health support. Results demonstrate the effectiveness of ML-based approaches for automated stress detection in educational environments.