Comparative Analysis of Machine Learning Algorithms for Detecting Suicidal Ideation in Social Media Content

Abstract

Suicidal ideation in students is one of the most serious threats that academic institutions have to deal with, as studies have shown that 10-24% of college/university students have thought of taking their own life seriously. Presently, all current intervention strategies practiced in campus guidance offices tend to be reactive crisis management only and do not seem to focus on preventing the emergence of suicidal thoughts. This limitation is addressed in the present study by proposing a Suicide Ideation Analysis and Monitoring (SIAM) System that is grounded on the use of Artificial Intelligence and Natural Language Processing to assess suicidal ideation in social media content at an early stage. The research work made use of a large dataset containing 232,074 posts in total, obtained from a Reddit-based community on SuicideWatch. Seven machine learning algorithms were applied together with 10-fold cross validation: Random Forest, Naive Bayes, Decision Tree, Support Vector Machine (SVM), K-Nearest Neighbor, Logistic Regression and Gradient Boosting. The models were evaluated using several performance metrics, including accuracy, precision, recall and F1 score. Overall, the SVM classifier was the best performing one as it produced 93% accuracy for every other metric, compared to the 90% accuracy of Random Forest and 92% accuracy of Logistic Regression. Also, support for Anthropic AI system provides an option for tagging and translating Filipino (Tagalog) and Cebuano to English which enables its possible use in schools and universities in the Philippines. The research demonstrates the potential of AI-driven solutions in transforming suicide prevention efforts from reactive to proactive approaches, while maintaining privacy considerations through careful handling of social media data. The developed system represents a significant advancement in campus mental health support systems, providing automated, privacy-conscious monitoring of suicide risk indicators that could enable earlier intervention and support for at-risk students.