Detecting Suicidal Ideation in Social Media Using Machine Learning and Ensemble Learning Technique

Suicide is one of the problems that is seriously discussed in all countries of the world. These days, since many people use social media, it has led to them posting their feelings, including suicidal thoughts, on these social networking sites. The aim of this paper is to use machine learning algorithms and ensemble learning technique (boosting) to identify suicidal thoughts from non-suicidal ones in social media posts. The dataset used in this study consists of users' posts on Twitter and Reddit that were combined manually to create a unified dataset. Various features were used for feature extraction and various machine learning algorithms and one ensemble learning technique are used for classifying suicidal and non-suicidal posts. This study shows that the ensemble learning method performed better than machine learning algorithms, in which AdaBoost achieved an overall accuracy of ۹۷.۳۶% and Cat Boost achieved accuracy of ۹۵.۶۱% and a cross-validation accuracy of ۹۸.۰۷%. The findings show that the use of various features and their right combination can supply better performance in identifying suicidal thoughts.

کلیدواژه ها:

Suicidal ideation ، Machine learning ، Ensemble learning ، Social media ، Dataset