Mental Health Prediction from Twitter Content Data based on Stress Inducing Themes and Quantum Neural Network Classifier

The increasing prevalence of social media provides novel approaches to understanding mental health states through digital interactions. This study introduces a Quantum Neural Network (QNN)-based framework for predicting the emotional states of mood-disordered patients by analyzing stress-induced themes and personality traits extracted from Twitter content. Stress themes, including negative self-reflection, exposure to negativity, and addictive behaviors, are identified utilizing BERTopic models, while both self-reported and automated personality traits are assessed through Sentence-BERT. The Authenticity Score quantifies discrepancies between self-reported and inferred personality traits, thereby enhancing the reliability of classification. The Quantum Neural Network (QNN) classifier effectively captures complex, non-linear interactions between stress induced themes and personality traits, which improves the performance of emotional state prediction for mood-disordered patients compared to traditional deep learning models. Hence, QNN achieves high precision (94%), recall (93%), F1-score (93.5%), and accuracy (94.5%) in identifying mental health risks. Furthermore, the study provides the importance of temporal features in monitoring mood variations and improving early diagnosis for conditions such as depression, anxiety, bipolar disorder, obsessive-compulsive disorder (OCD), post-traumatic stress disorder (PTSD), and schizophrenia.