Empirical Assessment and Detection of Suicide Related Posts in Twitter using Artificial Intelligence enabled Classification Logic

Abstract

The identification of suicidal thoughts in online social networks is an expanding field of study fraught with major challenges. Recent studies have shown that the readily available data, dispersed over many online life phases, contains useful clues for accurately identifying persons with suicidal intentions. The primary challenge in preventing suicide is learning to recognize and respond appropriately to the sometimes-confusing risk factors and warning indications that may precipitate an attempt. Indicators useful for diagnosing people with suicide thoughts can be found in publicly available material shared over social media platforms, according to recent studies. Understanding and recognizing the myriad risk factors and warning symptoms that may precede a suicide attempt is the primary difficulty in this area of public health. In this research, we developed a benchmark for multi-class categorization using machine learning models. We used a majority classifier, a frequency-based technique, and two deep learning models as our models. Both deep learning models outperformed the majority and the word frequency classifier, with results that were very comparable. These classification results are on par with the state-of-the-art on similar problems and, in most cases, with human results.