Deteksi Potensi Bunuh Diri Melalui Pesan Teks Menggunakan Model Deep Learning

Suicide is a serious problem that requires early detection efforts, one of which is through the analysis of text messages written by individuals on social media. Text as a communication medium has limitations in describing an individual's emotional state, due to the lack of nonverbal indicators such as voice and facial expressions. Therefore, this study aims to detect potential suicides through text messages using deep learning techniques. The Long Short-Term Memory (LSTM) model is used as the main architecture for classification, and Global Vectors for Word Representation (GloVe) is used as the word representation technique. The dataset used comes from the Reddit platform, chosen because of its numerous discussions related to mental health issues and potential suicide. Evaluation results show that the proposed model is able to achieve an accuracy level of 93%. This result shows an improvement compared to previous research, indicating that this approach is effective in detecting indications of potential suicide through text messages. This research is expected to contribute to the development of an artificial intelligence-based early detection system to support suicide prevention efforts.