Deep Belief Networks and Hybrid Term-Frequency Inverse Document Frequency (TF-IDF) Based Suicide Identification in Online Social Networks

Abstract

With the growing popularity of Online Social Networks (OSNs) like Facebook, Instagram, Twitter and YouTube, most individuals believe that writing about their emotions and sharing their preferences and views on social media is more convenient than doing it in person. These networks are improving associated to a dissimilar of societal events, such as illness, unhappiness, and even suicide. In this scenario, the cooperative of people who have suicidal thoughts (ST) and use OSNs rapid their intentions. Our method combines Deep Belief Networks (DBN), a type of deep learning model, with Hybrid Term-Frequency Inverse Document Frequency (HTF-IDF), a popular text mining method for collecting feature vectors from textual data. The DBNs are trained utilizing TF-IDF features extracted from user postings, permitting for the detection of subtle patterns and semantic subtleties that designate suicidal intent. The HTF-IDF component assistance in highlighting significant yet unusual terms across postings, which are important users’ emotional and psychological states. The project began with the collection of a dataset containing dissimilar kinds of communication from numerous main online social networks, which was then preprocessed account for the linguistic and syntactic characteristics of online communication. The proposed DBN-TF-ID model was tested against many benchmarks and outperformed existing machine learning algorithms in terms of accuracy and sensitivity in detecting probable suicidal material. To conclude, the use of DBN-HTF-IDF to examine social media material provides a viable path for developing suicide prevention tools, potentially leading to more prompt and effective interventions in mental health emergencies.