Early mental health risk detection using social media

This project focuses on detecting various mental health conditions by analyzing social media text data using advanced deep learning models. Social media has become a significant platform for individuals to express their thoughts, making it an ideal source of data for mental health monitoring. The approach involves utilizing three powerful deep learning algorithms: BERT combined with CNN, DistilBERT with BiLSTM, and DeBERTa with BiLSTM, which are employed to classify user-generated text into seven different mental health categories: Normal, Depression, Suicidal, Anxiety, Stress, Bipolar, and Personality Disorder.

The dataset used for training these models consists of various social media posts related to mental health, which are analyzed to identify subtle linguistic patterns and sentiments indicative of specific mental health conditions. These models are designed to understand complex language patterns and classify text effectively, offering insights into the user's mental health based on their written expression.

The primary goal of this project is to evaluate the performance of these models in classifying mental health conditions and to determine the most effective approach for this task. By leveraging the power of deep learning techniques, this study demonstrates the potential of using text analysis to monitor mental health. It highlights how machine learning models can be used to understand the mental state of individuals from their social media posts, which can serve as an early indicator for potential mental health concerns.

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