Detection of Suicidal Ideation on Social Media Profiles Using Machine Learning Techniques

Abstract

Nowadays, due to numerous reasons like the nuclear family, peer pressure for fake prestige, impatience mindset, and mental pressure has become a usual trait in every person. When someone finishes their own life, we say that they passed by suicide. A suicide attempt denotes that someone tried to end their life, but did not die. It has been a crucial issue in current society. For this earlier detection and prevention of suicide attempts should be handled to protect people’s life. Today with the expansion of technology people tend to express their emotions on social media. A massive amount of people vents out their emotions online as they have no support system in actual life. It has been noticed or seen a lot of times those suicidal trends varying from mild to an extreme could be from a person's online profile activity. In this article, we put ourselves in a tough context, on the opinions that could be thinking of suicide. Particularly, we propose to address the shortage of terminological resources connected to suicide by a technique of assembling a vocabulary associated with suicide. After that, we proposed a specific method that includes all critical criteria which could be demonstrated by a suicidal person using Natural Language Processing (NLP) techniques. Our approach indicates efficiently an actual, mentally worried profile from a typical profile. Finally, we summarized to encourage future research. We also summarized the limitations of existing work and provide an outlook of further research approaches. This study provides an explanation as well as a solution by classifying the Reddit suicide and non-suicide opinion using various algorithm. Among these algorithms, Logistic Regression accuracy is the best accuracy that is 92.97%. The proposed model is made on Jupyter Notebook (a Python-based IDE) and trained on Kaggle's standard Suicide and depression dataset which has 2,33,338 records.