Skip to content
Open access

A Comparative Evaluation of Machine Learning and Deep Learning Models for Detecting Depressive behavior on X

2026 · International journal of research and innovation in social science · 0 citations

Abstract

Early detection of mental health concerns through social media platforms like X offers significant potential for proactive intervention. However, identifying nuanced emotional and psychological distress from short, informal texts remains a complex challenge. This study aims to bridge this gap by developing and comprehensively evaluating a spectrum of machine learning, deep learning, and transformer-based models for detecting depressive behavior from X posts. A robust natural language processing (NLP) pipeline was established, encompassing data cleaning, text preprocessing, feature extraction, and classification. Four distinct architectures were investigated: Logistic Regression, Linear Support Vector Machine (SVM), CNN-LSTM, and Fine-Tuned Distil BERT. Traditional models utilized TF-IDF representations, while deep learning and transformer models employed token-based representations. Model performance was rigorously benchmarked using accuracy, precision, recall, and F1-score. Experimental results reveal that all models achieved high efficacy with accuracies exceeding 96%. Notably, the Fine-Tuned Distil BERT secured the best overall performance with an accuracy of 97.46% and an F1-score of 96.68%, whereas Linear SVM demonstrated exceptional sensitivity with the highest recall of 97.05%. These findings demonstrate that transformer-based models and advanced NLP pipelines significantly enhance the reliability of automated mental health screening on social media, providing a scalable foundation for intelligent, text-based psychological monitoring systems.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.