Tools
Tools
Anora, Anirudh (2024) Depression detection: Comparative Analysis of different AI models trained on data threads from Reddit. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Depression is amongst the most common mental diseases which affect millions of people all over the world and present a number of issues for society and healthcare systems. Earlier methods involved didactic structured interviews and self-administered questioners, but with improvements in NLP and AI we now have options for automated recognition. Through this research, the author seeks to understand how each of the proposed deep learning models can be applied to accurately detect signs of depression from text data mined from Reddit – a social media forum in which users post about numerous conditions including mental health. The models that we consider in our analysis are LSTM, GRU, BiLSTM, BERT, and FFNN as the baseline. The data collection includes samples that are tag lassen as either “depression” or “non depression.” As a part of preprocessing, text cleaning was performed by applying features of tokenization, stop-word removal, stemming/lemmatization and numerical concept. These results indicate that, despite the fact that all models have potential, BERT is especially effective as it has a bidirectional context awareness of text meaning. FFNN is much more straightforward but can offer a sound benchmark. This work focuses on crucial aspects of using social media data and deep learning approach for the identification of depression, which could present a pragmatic, cost-effective and privacy-preserving solution for mental health care and support for practitioners, as well as contribute to the optimization of outcomes for patients.
Actions (login required)
 |
View Item |