Abstract
A significant proportion of suicide attempts, especially among adolescents and young adults, remain unacknowledged despite the diligent efforts of suicide prevention teams and extensive research. Due to the expressive nature of social media, individuals might reveal aspects of their inner lives, potentially indicating subtle indicators of mental health disorders. This study examines how data gathered from social media platforms can enhance the identification of those at heightened risk of contemplating or attempting suicide. The proposed model's structured training methodology comprises three steps: data preparation, feature selection, and model training. Data preparation constitutes a substantial obstacle in the KDD process, and feature selection is essential for deriving a concise yet representative sample that encapsulates relevant suicide-related outcomes. Feature selection is essential for enhancing the performance of the TriBranch CNN model utilised in the study. Experimental results indicate that the proposed technique significantly outperforms state-of-the-art models such as BiLSTM and traditional CNNs. The system demonstrates significant potential in identifying suicide attempts among young adults, achieving an accuracy of 95.72 percent. Given these findings, mental health programs ought to integrate social media analytics, which can serve as an effective instrument for suicide prevention teams in the early detection and intervention phases.