Mental Health Safety and Depression Detection in Social Media Text Data: A Classification Approach Based on a Deep Learning Model

This project focuses on identifying mental health status, specifically detecting depression, through analyzing social media text data using deep learning models. The approach uses four advanced algorithms: BERT combined with CNN, DistilBERT with BiLSTM, DeBERTa with BiLSTM, and DistilGPT2 paired with BiLSTM. These models classify user-generated text as either healthy or depressed. The dataset consists of various social media posts related to mental health. The system is designed with modules for user registration, login, sentence classification, and logout. A web application is developed using Flask as the backend framework and HTML, CSS, and JavaScript for the frontend. This setup enables users to interact with the classification system efficiently. The project aims to contribute to mental health monitoring by providing a reliable automatic detection tool based on text data analysis. The models are evaluated to find the best performing method for this classification task. This study highlights the potential of deep learning techniques in understanding mental health signals from text inputs.

Keywords: Mental health, depression detection, social media, text classification, deep learning, BERT, CNN, BiLSTM, DistilBERT, Flask.