Comparative Performance Analysis of Ensemble and Attention-Based Deep Learning Methods for Depression Classification

Nur Sultan Yüce, Abdullah Ammar Karcioğlu, Mesut Karabacak

2025

Abstract

Depression is a globally prevalent psychological disorder that significantly impairs individuals' quality of life. Early diagnosis and timely intervention are essential for effective treatment and societal reintegration. This study conducts a comparative performance analysis of ensemble learning methods including XGBoost, Random Forest, LightGBM, Gradient Boosting Machine (GBM), and CatBoost and deep learning models such as Deep Neural Networks (DNN) and TabNet for depression prediction. Using a publicly available dataset, we applied various preprocessing and hyperparameter optimization techniques to enhance model performance and mitigate overfitting. Experimental results demonstrate that the LightGBM model achieves the highest classification accuracy (92.77%) and ROC-AUC (0.976), outperforming other models. These findings indicate that ensemble-based approaches are highly effective for early depression detection and hold promise for integration into data-driven clinical decision support systems

Paper Citation

in Harvard Style

Yüce N., Karcioğlu A. and Karabacak M. (2025). Comparative Performance Analysis of Ensemble and Attention-Based Deep Learning Methods for Depression Classification. In Proceedings of the 2nd International Conference on Advances in Electrical, Electronics, Energy, and Computer Sciences - Volume 1: ICEEECS; ISBN 978-989-758-783-2, SciTePress, pages 100-105. DOI: 10.5220/0014363500004848

in Bibtex Style

@conference{iceeecs25,
author={Nur Sultan Yüce and Abdullah Ammar Karcioğlu and Mesut Karabacak},
title={Comparative Performance Analysis of Ensemble and Attention-Based Deep Learning Methods for Depression Classification},
booktitle={Proceedings of the 2nd International Conference on Advances in Electrical, Electronics, Energy, and Computer Sciences - Volume 1: ICEEECS},
year={2025},
pages={100-105},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0014363500004848},
isbn={978-989-758-783-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Advances in Electrical, Electronics, Energy, and Computer Sciences - Volume 1: ICEEECS
TI - Comparative Performance Analysis of Ensemble and Attention-Based Deep Learning Methods for Depression Classification
SN - 978-989-758-783-2
AU - Yüce N.
AU - Karcioğlu A.
AU - Karabacak M.
PY - 2025
SP - 100
EP - 105
DO - 10.5220/0014363500004848
PB - SciTePress