A DEEP LEARNING-BASED HYBRID MODEL FOR AUTOMATICALLY DETECTING DEPRESSION IN SOCIAL MEDIA POSTS

In recent times, mental health issues have been on the rise, influenced by various factors, including lifestyle changes. With the widespread use of social media, individuals from different backgrounds can openly share their thoughts and emotions, providing valuable data for research. This has opened the opportunity to analyse social media discussions to evaluate the potential presence of depression by examining the sentiments conveyed in the text. Although various heuristic methods for depression detection (DD) are available, the rise of AI has enabled the development of more efficient learning-based methods. However, since a single approach may not be universally applicable, there is a need to refine DL models to improve their performance in detecting depression. In this paper, we introduce a DL context that integrates CNN and BiLSTM networks, enabling the model to capture both features from the data and temporal dependencies. We present an algorithm called Learning Based Depression Detection (LBDD), which analyses Twitter posts to classify them based on the probability of depression. After evaluating the approach on a standard dataset, the projected model outclasses several prevailing methods, achieving an accuracy of 96.32%.