Detecting and Analyzing Suicidal Ideation on Social Media Using Deep Learning and Machine Learning Models.

Area of Science:

• Computational linguistics
• Artificial intelligence
• Mental health informatics

Background:

• Suicidal ideation is often expressed on social media, presenting an opportunity for early detection.
• Analyzing social media for suicidal ideation patterns is complex.
• Automated systems are needed to identify behavioral changes indicative of suicidal ideation.

Purpose of the Study:

• To develop and evaluate a machine learning system for automated early detection of suicidal ideation.
• To compare the performance of different machine learning models and feature sets for this task.

Main Methods:

• Utilized publicly available Reddit datasets.
• Employed TF-IDF and Word2Vec for text representation.
• Implemented hybrid deep learning (CNN-BiLSTM) and machine learning (XGBoost) models for classification.
• Conducted experiments using textual and LIWC-22 features.

Main Results:

• The CNN-BiLSTM model achieved 95% accuracy in detecting suicidal ideation using textual features.
• The XGBoost model achieved 91.5% accuracy using textual features.
• XGBoost outperformed CNN-BiLSTM when utilizing LIWC features.

Conclusions:

• Machine learning models, particularly CNN-BiLSTM with textual features, show high potential for accurate suicidal ideation detection.
• The choice of features (textual vs. LIWC) impacts model performance, suggesting hybrid approaches may be beneficial.
• Automated analysis of social media offers a promising avenue for early intervention in mental health crises.