Suicide Ideation Detection using Multiple Feature Analysis from Twitter Data

Abstract

According to the World Health Organization, over 700,000 people die by suicide each year, with many more attempting suicide. Two of the most effective approaches to prevent suicide attempts are early recognition of suicidal ideation and proper treatment. People suffering from despair and suicidal ideation are increasingly turning to social media to share their thoughts. According to researchers, there is a clear relationship between an individual's mental health and their associated language use. The primary goal of the research is to examine online Twitter tweets and discover the features that may signal suicide ideation in users. To train the data and assess the performance of the proposed method, machine learning and natural language processing techniques were used. Several Linguistic, Topic, Temporal Sentiment and Statistical features are extracted and combined to achieve an accuracy of 87% using Logistic Regression classifier. According to the research, proper selection of features and their combinations help in achieving better performance.