An Innovative Method to Predict Suicide Attempts from Social Media Based on BP-RBM Approach

Suicide is one of the top 10 causes of death worldwide, with each occurrence significantly impacting almost 140 individuals. This concerning number highlights the imperative of creating efficient ways to avert similar incidents. Social media has a significant volume of user-generated content, allowing for the detection of persons at danger of suicide and the creation of prompt, data-driven interventions to support them. This study examines the viability of identifying such individuals via the analysis of social media data. The textual data is first cleaned to ensure machine interpretability and suitability for ML models. TF-IDF vectorization and FastText embedding are two prominent methods utilized for feature extraction. They created a BP-RBM model utilizing carefully chosen variables to improve its predictive accuracy. It assessed the model's efficacy in comparison to conventional BP and RBM techniques. The results demonstrate that the suggested BP-RBM model surpasses both baseline models, with an impressive accuracy rate of 97.45 % in identifying individuals on social media who may be at risk of suicide. The findings illustrate the potential effectiveness of combining machine learning with social media analysis for proactive mental health monitoring. The research reveals an effective approach for identifying and aiding individuals before to self-harm, thereby employing technology to avert suicides.