Suicide risk prediction for users with depression in question answering communities: A design based on deep learning

Predicting suicide risk for people with depression is crucial for preventing adverse events. Existing research has mainly focused on users’ current post in online communities, overlooking historical posts that can provide a comprehensive representation of users’ emotional changes. In this study, we propose a deep learning-based method called dynamic historical information-based suicide risk prediction (DHISRP), which integrates current and heterogeneous historical posts to capture the dynamic and complicated features of users’ post sequences for suicide risk prediction. Empirical evaluation shows the superior effectiveness of our method compared to the baseline model and emphasizes the importance of considering both current and historical posts to predict suicide risk.