Jul 2026· International journal of computer information systems and industrial management applications· Vol 18, pp. 1174-1188· 0 citations
TL;DR
The results demonstrate that the integration of rich contextual embeddings from BERT with transparent XAI produces a trustworthy, accurate, and actionable system for mental health risk monitoring.
Abstract
Mental health disorders represent a growing global public health concern, yet early detection remains challenging due to the reliance on conventional clinical assessments. The rise of social media provides a unique opportunity to monitor individuals' emotional states through their textual expressions passively. However, existing automated detection approaches are predominantly opaque "black-box" systems, limiting their adoption in clinical and educational settings where interpretability is paramount. This paper presents a comprehensive, multi-modal pipeline for automated mental health risk detection that integrates contextual embeddings from BERT (Bidirectional Encoder Representations from Transformers) with behavioral and linguistic indicators. We benchmark five distinct classifier architectures, including TF-IDF-based and BERT-based models, using Stratified 5-Fold Cross-Validation across multiple metrics: Accuracy, Macro F1-Score, Cohen's Kappa (κ), and Matthews Correlation Coefficient (MCC). The best-performing model, BERT + Logistic Regression, achieves a Macro F1-Score of 0.6464 ± 0.0119 and a Macro AUC-ROC of 0.8726, while BERT + Random Forest achieves the highest accuracy of 0.8742 ± 0.0020. To address the interpretability gap, we implement a multi-faceted Explainable AI (XAI) framework comprising Shapley Additive exPlanations (SHAP) for global and local feature attribution, Local Interpretable Model-agnostic Explanations (LIME) for word-level insights, and Counterfactual Analysis for decision-boundary understanding. Statistical significance is rigorously validated via McNamara’s test(p <0.001) and the Wilcoxon signed-rank test. Our results demonstrate that the integration of rich contextual embeddings with transparent XAI produces a trustworthy, accurate, and actionable system for mental health risk monitoring.
Mental health disorders such as anxiety, depression, and Mild Cognitive Impairment (MCI) are increasingly prevalent among young adults aged 18–30, significantly affecting academic performance, cognitive functioning, and overall wellbeing. Traditional diagnostic approaches depend on subjective assessments and limited clinical observations, making early and accurate detection challenging. To address these limitations, this research proposes an interpretable deep learning-based multimodal method for comprehensive mental health prediction and personalized intervention. The framework integrates heterogeneous data sources, including demographic, cognitive, behavioral, physiological, and neurocognitive indicators collected from clinical settings. Data preprocessing includes imputation, normalization, encoding, and text transformation. A Cross-Directional Feature Learning Network (CDFLN) is employed for robust multimodal feature extraction, followed by a Multi-model Progressive Dense Self-Attention for Cross Domain (MPDSA-CD) architecture for classification of anxiety, depression, and MCI, along with cognitive risk and severity assessment. Model performance is further enhanced by the Starfish Optimization Algorithm for hyperparameter tuning and parameter refinement. To ensure clinical transparency, SHapley Additive exPlanations (SHAP) are utilized to interpret model predictions and identify key risk factors influencing mental health outcomes. The proposed method achieves an accuracy of 99.8%, precision of 99.7%, recall of 99.9%, and F1-score of 99.8%, demonstrating strong robustness, generalization ability, and clinical applicability for early detection and effective psychological intervention in young adults.
Rupali D. Kasar, Garima Shukla· Frontiers in Signal Processi...· 0 citations
Mental health disorders such as depression, anxiety, and post-traumatic stress disorder (PTSD) affect over one billion
people worldwide, yet early detection remains a major clinical challenge. In recent years, text data from social media posts,
clinical notes, and patient surveys has emerged as a rich source of signals for automated mental health screening. However,
most existing machine learning models operate as black boxes, limiting clinical adoption. This paper presents an interpretable
machine learning framework that combines natural language processing (NLP) feature extraction with explainable AI
techniques — specifically SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic
Explanations) — to predict mental health conditions from text while providing transparent, clinically meaningful
explanations. A multi- class classification task involving depression, anxiety, PTSD, and healthy controls is performed on a
dataset of 19,320 labelled text samples. The proposed XGBoost model with SHAP explanations achieves 87.3% accuracy and
an AUC of 0.924, while the fine-tuned BERT model achieves 91.6% accuracy and an AUC of 0.961. Experimental results
demonstrate that interpretability does not significantly compromise predictive performance, enabling trustworthy AI-assisted
mental health screening.
N. Thakur, D. Patil, Pushpa Choudhary· International Journal for Re...· 0 citations
Maternal mental health disorders remain a major public health problem, particularly in low- and middle-income countries, where limited mental health resources and inadequate screening systems lead to late diagnosis and intervention. Recent advances in artificial intelligence have demonstrated great potential in improving the prediction of early maternal mental health risks. However, most current methods are limited to traditional machine learning models and lack interpretability. This study introduces the explainable Hybrid Classical-Quantum Support Vector Machine (SVM) framework for the prediction of maternal mental health risk using multicountry psychosocial and clinical data from Uganda and Pakistan. The proposed framework integrates classical SVM learning with quantum-enhanced feature representations, while SHapley Additive exPlanations (SHAP) are leveraged to increase the model transparency and clinical interpretability. A comparative evaluation was performed against Random Forest, XGBoost, LightGBM, CatBoost, Stacking Ensemble, Classical SVM, and standalone Quantum SVM using Accuracy, Precision, Recall, F1-score, Receiver Operating Characteristic Area Under the Curve (ROC-AUC) and Precision-Recall Area Under the Curve (PR-AUC). We performed two experimental settings: one using the full set of features and another using a set of features without leakage, i.e. excluding variables that directly contribute to the EPDS-derived outcome. Under the complete feature set, the proposed Hybrid Classical-Quantum SVM achieved 99.86% accuracy, 99.80% F1-score, 99.92% ROC-AUC and 99.91% PR-AUC, while maintaining competitive performance after the leakage-free evaluation. The SHAP analysis provided enhanced model interpretability, identifying clinically meaningful demographic, obstetric, socioeconomic and psychosocial factors associated with maternal mental health risk. Our results demonstrate that hybrid classical-quantum learning offers a competitive and explainable paradigm for predicting maternal mental health risk and highlights the complementary nature of quantum feature representations in healthcare analytics.
Shallon Ahimbisibwe, Emmanuel Ahishakiye, S. Maling et al.· Discover Artificial Intellig...· 0 citations
Mental health disorders are frequently under-diagnosed because clinical decisions rely on isolated data sources such as structured screening scores or unstructured self-reported narratives, rarely both together. This paper proposes a multimodal mental health prediction framework that jointly models structured clinical and lifestyle attributes (e.g., PHQ-9 and GAD-7 scores, sleep duration, physical activity, demographic variables) together with unstructured free-text self-statements collected from the same individual. The proposed architecture combines a collaborative cross-modal learning module, which aligns tabular and textual representations in a shared latent space via co-attention, with a conditional Generative Adversarial Network (GAN) that synthesizes minority-class embeddings to counter the class imbalance typical of mental health datasets. A simulated multimodal cohort of 5,000 respondents, each contributing paired structured responses and a free-text statement, is used to train and evaluate the framework. Experimental results show that the proposed Collaborative-GAN model achieves 91.4% accuracy and an AUC-ROC of 0.95, outperforming unimodal and naive late-fusion baselines by 4 to 9 percentage points across all metrics. These findings suggest that collaborative multimodal fusion combined with adversarial data augmentation can meaningfully improve early mental health risk screening, and the framework is presented as a reproducible template that researchers can adapt to real clinical datasets.
Gopinath, Yogish Naik· Journal of Intelligent Decis...· 0 citations
It is shown that con-textual transformer/LLM models yield more reliable macro-level performance under imbalance than TF–IDF baselines, particularly for semantically adja- cent classes.
Nehal Shah, Mehul Barot· International journal of com...· 0 citations
To address the susceptibility of pre-trained language models (PLMs) to spurious correlations in mental health analysis, we propose a robust framework designed to disentangle genuine clinical signals from superficial artifacts. Our approach integrates a contrastive learning method by using masked language modeling (MLM) with artifact-masked data and an adaptive hard-negative sampling strategy to ensure cross-domain generalization. Extensive evaluations on diverse benchmarks, including real-world social media posts and LLM-generated synthetic conversational utterance datasets, demonstrate that our method significantly outperforms baselines in out-of-distribution (OOD) settings, particularly as contextual information increases. Importantly, the learned representations capture clinically meaningful patterns consistent with well-documented psychiatric comorbidities (e.g., depression-eating disorder), suggesting genuine clinical signal capture rather than superficial pattern matching. These findings demonstrate the effectiveness of our framework in detecting mental disorders in real-world settings, suggesting its practical applicability in social media platforms and conversational AI systems.
Dongik Shin, Youngwook Ham, Sang Won Lee et al.· Proceedings of the 32nd ACM...· 0 citations
