Abstract
Mental health disorders such as depression, anxiety, and stress are rising globally, necessitating effective and scalable monitoring solutions. Social media platforms provide an abundant source of real-time behavioral and linguistic data, which can be leveraged for early detection of mental health conditions. However, most existing methodologies rely solely on textual features and traditional machine learning or deep learning models, limiting their ability to capture the multi-faceted nature of user behavior and interactions. These approaches often lack interpretability and fail to generalize well across diverse social contexts, reducing their practical applicability. To address these limitations, this paper proposes MindMonitorAI, an AI-driven framework for mental health analysis using social media data. The framework introduces a novel hybrid deep learning model, MindFusionNet, which integrates three complementary modalities: textual posts, behavioral engagement metrics, and social graph connections. The architecture employs BERT-based encoders for text analysis, neural networks for behavioral features, and graph neural networks for modeling social interactions, followed by attention-based multi-modal feature fusion. Explainability techniques, including SHAP and LIME, are incorporated to enhance transparency and support ethical decision-making. An experimental evaluation of multiple publicly available mental health datasets demonstrates that MindFusionNet achieves superior performance, attaining an accuracy of 98.68%, outperforming baseline machine learning and deep learning models. The results validate the efficacy of multi-modal integration and attention mechanisms in improving predictive capabilities. The proposed framework is significant for real-time, interpretable mental health monitoring systems, providing actionable insights for healthcare professionals and supporting timely intervention strategies.
Keywords: Behavioral Analysis, Explainable AI, Mental Health Analysis, Multi-Modal Deep Learning, Social Media Monitoring