Abstract:
Stress and anxiety can have profound impacts on an individual's mental and physical well-being,
making their early detection crucial for timely intervention and support. This study investigates
the application of machine learning for detecting stress and anxiety levels from survey data, which
is crucial for timely intervention and support. A comprehensive experimental analysis evaluated
various algorithms, including Support Vector Machines (SVM), XGBoost, CatBoost, ensemble
methods like Random Forest, and traditional techniques like Logistic Regression. The
experimental setup utilized powerful computational resources and employed tools like NumPy,
Pandas, and Scikit-learn. Extensive data preprocessing and feature engineering ensured data
quality. Results revealed SVM, XGBoost, and CatBoost as top performers, achieving highest
accuracy and F1 scores, demonstrating effectiveness in capturing complex stress and anxiety
patterns. However, challenges like data quality, feature selection, individual variability, and
interpretability were encountered. The study highlights machine learning's potential for stress and
anxiety detection, providing insights into algorithm selection and performance trade-offs. Future
work may focus on incorporating domain knowledge, addressing ethical considerations, and
deploying the model for improved mental health support.