XplainMentalFusion: 可解释的多类别精神疾病预测框架

Early and dependable recognition of mental health disorders, which rank among the largest contributors to worldwide disease burden, is central to timely care. Text-based screening tools can assist clinicians, yet most published approaches handle only two classes, draw on a single data source, struggle with skewed class distributions and give clinicians little insight into why a decision was made. We present XplainMentalFusion, an interpretable ensemble of deep learning models that labels each text as neutral or as one of five conditions (anxiety, depression, bipolar disorder, schizophrenia and