KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 7 Serial Number 44 2026 04 10 Deep Neural Network Modeling of Adolescent Generalized Anxiety Using Attentional Bias, Physiological Arousal, Worry Severity, and Parent–Child Conflict Deep Neural Network Modeling of Adolescent Generalized Anxiety Using Attentional Bias, Physiological Arousal, Worry Severity, and Parent–Child Conflict 1 11 10.61838/kman.jayps.5235 EN Journal Article 2025 10 23 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> To develop and evaluate a Deep Neural Network model capable of accurately predicting generalized anxiety in adolescents by integrating multimodal cognitive, neurocognitive, physiological, and environmental features.</p> <p><strong>Methods and Materials: </strong>This cross-sectional study included adolescents from Chile. Participants completed a comprehensive, multimodal assessment protocol comprising the Generalized Anxiety Disorder Assessment- &nbsp;(GAD- ) to classify anxiety severity, the Penn State Worry Questionnaire for Children to measure conscious worry severity, and the Conflict Behavior Questionnaire to assess parent-child conflict. Implicit neurocognitive attentional bias to threat was quantified using a computerized dot-probe task. Concurrently, continuous electrocardiography and electrodermal activity recordings were utilized to capture objective physiological arousal, specifically heart rate variability and skin conductance responses. A Deep Neural Network architecture was subsequently constructed, trained, and tested to model the complex, non-linear relationships among these disparate variables and predict adolescent generalized anxiety classification.</p> <p><strong>Findings: </strong>The trained Deep Neural Network demonstrated exceptional predictive performance, achieving an overall classification accuracy of and an Area Under the Receiver Operating Characteristic Curve of on unseen testing data. Feature importance analysis derived from the network’s weights revealed that physiological arousal constituted the largest broad predictive domain ( &nbsp;total; comprising for heart rate variability and for skin conductance responses). Conscious worry severity emerged as the strongest individual predictor, accounting for of the model’s predictive capacity. Furthermore, environmental and neurocognitive factors contributed significantly to the predictive architecture, with parent-child conflict accounting for and attentional bias to threat accounting for of the model’s weight.</p> <p><strong>Conclusion: </strong>Multimodal deep learning architectures provide a highly accurate and comprehensive framework for predicting adolescent generalized anxiety, highlighting the critical, synergistic interplay of somatic hyperarousal, severe cognitive worry, and familial discord.</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/jayps/article/download/5235/9508