Leveraging Hybrid CNN-LSTM Model with Attention Mechanisms for Detection of Mental Health of Patients Using Twitter Data

Abstract

Depression, affecting over 280 million people globally, poses a significant public health challenge due to delayed diagnoses and limited access to traditional mental health assessments. This study proposes a hybrid deep learning model combining Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM) networks with a Multi-Head Attention mechanism to detect depressive linguistic markers in Twitter data. Using a curated dataset of tweets collected via the Twitter API (2018-2023), the model leverages advanced preprocessing with the Isolation Forest algorithm, SMOTE for class balancing, and hyperparameter tuning to achieve an accuracy of 98.66%, precision of 98.86%, recall of 98.52%, and an Fl-score of 98.69%. Integrated with Explainable AI (XAI) via LIME, the model provides transparent, word-level insights into predictions. [cite_start]Deployed through a Flask web application with a Tailwind CSS interface, it offers real-time depression detection, making it a scalable, interpretable tool for mental health monitoring and early intervention. [cite: 7, 8, 9, 10, 11]

Keywords

References