Depression Classification using Bert Embedding Model on Social Media Posts

Developing sophisticated methods to precisely detect health-related concerns on social media, including identifying sadness and anxiety, has become essential due to the expansion of the Internet. These systems focus on using machine learning techniques to determine the meaning and structure of writings shared by users on social media. Social media users’ data is confusing and inconsistent. Novel methods using deep learning and social networking platforms data to detect health issues. Provide just a little bit of information and understanding on the various texts individuals provide. This investigation introduces an innovative approach utilizing BERT to accurately and specifically detect posts related to sadness and anxiety. This approach preserves the contextual and semantic importance of words across the collection. The researcher employed word2vec, fasttext, BERT and Enhanced Grey Wolf Optimizer (E-GWO) and Deep Learning technologies to promptly analyze and detect indications of anxiety and melancholy in social media messages. Our solution surpasses previous advanced methods and incorporates the knowledge distillation methodology to achieve an accuracy of 95.9%.