A Social Media Intelligence Approach to Predict Suicidal Ideation from Sinhala Facebook Posts

The global increase in mental health issues, including suicidal ideation and behavior, has become a significant concern. Social media platforms like Facebook have provided individuals with an outlet to express their thoughts and emotions, creating an opportunity for mental health professionals and researchers to gain valuable insights into their mental wellbeing. As a result, sentiment analysis has emerged as a powerful tool for extracting sentiments from textual data, offering a promising avenue for predicting suicidal tendencies through social media posts. While previous studies have explored sentiment analysis for this purpose in various other languages, a critical gap exists in research literature focusing on the Sinhala language, spoken by a substantial population in Sri Lanka and other regions in the world. Understanding the sentiments and emotional states of individuals expressed themselves in Sinhala on social media can have significant implications for suicide prevention effort. Machine learning algorithms and techniques can be leveraged to analyze and predict the likelihood of suicidal tendencies based on sentiments expressed in posts. In this paper, we therefore present a social media intelligence approach to predict suicidal ideation from Sinhala Facebook posts. The proposed predictive model provides invaluable insights to mental health professionals, facilitating the development of targeted interventions and support systems for individuals at risk. The proposed hybrid sentiment analysis methodology consists of a machine learning model and a sentiment lexicon trained on carefully selected labelled social media text data. Each text is then classified as either suicidal or not, using Naïve Bayes, Support Vector Machine, and Logistic Regression algorithms. When compared their results, we could achieve the best results with the Naïve Bayes algorithm resulting in an accuracy of 79% proving that social media intelligence can be used to predict suicidal ideation from Sinhala Facebook Posts.