A Multimodal Deep Learning Framework for Mental Health Detection Using Social Media Data

<title>Abstract</title> <p>Mental health disorders like stress, depression, and anxiety have become leading health problems around the globe, but traditional diagnosis techniques have issues related to late detection, lack of accessibility, and stigma. With an increase in the number of social networking sites and other social media, there is a chance of accessing important textual, visual, and behavioral data about early detection of mental health. In this regard, this paper presents a multimodal deep learning approach for the detection of mental health disorders with the help of social media data. This approach uses feature fusion for integrating language, visual, and behavioral data to enhance accuracy. Deep learning models are used to learn from the multimodal data for early detection of any mental health disorder. Results show that the proposed system performs better than the traditional machine learning systems and unimodal approaches in terms of accuracy, precision, recall, and F1-score.</p>