Detection of Suicidal Risk on Social Media: A Hybrid Model

2025-05-26

AI Summary

Addressing challenges in early suicide risk detection on social media—including difficulty in fine-grained classification, severe class imbalance, and low signal-to-noise ratio—this paper proposes a four-level automated suicide risk classification method for Reddit posts. We introduce a novel RoBERTa-TF-IDF-PCA hybrid architecture that jointly models deep semantic representations and interpretable statistical features. To mitigate decision bias toward underrepresented high-risk classes, we integrate SMOTE-based oversampling with back-translation data augmentation. Under rigorous evaluation, our model achieves a weighted F1-score of 0.7512—outperforming a pure RoBERTa baseline by 4.3% and significantly surpassing BERT, SVM, and XGBoost baselines across accuracy, recall, and class-balanced performance. The approach advances robust, interpretable, fine-grained psychological risk monitoring, establishing a new paradigm for clinical-grade mental health surveillance in noisy, imbalanced social media data.

Technology Category

• Natural Language Processing: Safety and Robustness
• Machine Learning: Dimensionality Reduction/Feature Selection
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Suicidal thoughts and behaviors are increasingly recognized as a critical societal concern, highlighting the urgent need for effective tools to enable early detection of suicidal risk. In this work, we develop robust machine learning models that leverage Reddit posts to automatically classify them into four distinct levels of suicide risk severity. We frame this as a multi-class classification task and propose a RoBERTa-TF-IDF-PCA Hybrid model, integrating the deep contextual embeddings from Robustly Optimized BERT Approach (RoBERTa), a state-of-the-art deep learning transformer model, with the statistical term-weighting of TF-IDF, further compressed with PCA, to boost the accuracy and reliability of suicide risk assessment. To address data imbalance and overfitting, we explore various data resampling techniques and data augmentation strategies to enhance model generalization. Additionally, we compare our model's performance against that of using RoBERTa only, the BERT model and other traditional machine learning classifiers. Experimental results demonstrate that the hybrid model can achieve improved performance, giving a best weighted $F_{1}$ score of 0.7512.

Problem

• Detect suicidal risk levels from social media posts
• Develop hybrid model combining RoBERTa and TF-IDF-PCA
• Address data imbalance and overfitting in classification

Innovation

• Hybrid RoBERTa-TF-IDF-PCA model
• Data resampling and augmentation techniques
• Multi-class suicide risk classification