Dwarf Updated Pelican Optimization Algorithm for Depression and Suicide Detection from Social Media.

Area of Science:

• Computational linguistics
• Mental health informatics
• Machine learning

Background:

• Depression and suicidal ideation are critical global health issues.
• Clinical assessments for these conditions face accessibility and stigma barriers.
• Existing computational methods struggle with data variability and model generalization.

Purpose of the Study:

• To develop a novel Suicide and Depression detection from Social Media (SADDSM) approach.
• To address challenges in data variability and model generalization in mental health detection.
• To improve the accuracy and reliability of automated mental health monitoring.

Main Methods:

• Data preprocessing included stop word removal, tokenization, and stemming.
• Feature extraction utilized TF-IDF, style features, and enhanced word2vec.
• A modified mutual information score was employed for feature fusion.
• An ensemble model of RNN, DBN, and improved LSTM was developed.
• The Dwarf Updated Pelican Optimization Algorithm (DU-POA) optimized model weights.

Main Results:

• The SADDSM model achieved a 0.962 accuracy rate with 90% training data.
• The ensemble deep learning approach demonstrated superior performance over existing techniques.
• DU-POA effectively fine-tuned model weights for enhanced predictive power.

Conclusions:

• The proposed SADDSM method offers a robust and accurate solution for detecting depression and suicidal thoughts from social media.
• The integration of advanced feature extraction, ensemble deep learning, and optimization algorithms significantly improves detection capabilities.
• This research highlights the potential of social media analysis for scalable and accessible mental health monitoring.