A Comparative Study of Machine Learning Models for Stress Detection in Academic Social Media Posts
This study evaluates machine learning approaches for stress detection in academic social media text using the Dreaddit dataset. Five classifiers were assessed across multiple feature configurations. A weighted ensemble with VADER sentiment features achieved 77.90% accuracy and 79.48% F1-Score on 715 test posts.
Authors
- Imama Tufail
Institutions
- Bahria University (PK)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-03-16
- DOI
- https://doi.org/10.5281/zenodo.19042726
- Primary Topic
- Mental Health via Writing
- Type
- article
- Field-Weighted Citation Impact
- 0.00