DETECTION OF SUICIDAL TENDENCIES IN REDDIT POSTS USING MACHINE LEARNING

Suicide remains a significant public health challenge, with early detection of suicidal tendencies being crucial for intervention and prevention. Online platforms like Reddit offer a unique opportunity to identify individuals expressing suicidal ideation through user-generated content. This study leverages machine learning techniques to detect suicidal ideation in posts from the "SuicideWatch" and "depression" subreddits, supplemented with non-suicidal posts from "r/teenagers." By utilizing a deep learning model that incorporates Bidirectional Long Short-Term Memory (BiLSTM) and an attention mechanism, we aim to enhance the accuracy and reliability of detection systems. The model achieved exceptional performance, attaining an AUC-ROC score of 0.98 on the validation set and 0.989 on the test set. This study underscores the potential of automated systems in identifying at-risk individuals, thereby facilitating timely interventions and supporting suicide prevention efforts. Early detection of suicidal tendencies in online communications can significantly improve mental health outcomes and save lives.