KMAN Publication Inc. (KMANPUB) Journal of Assessment and Research in Applied Counseling (JARAC) 8 Serial Number 28 2026 01 01 A Machine-Learning Approach to Modeling Cognitive Flexibility, Stress Reactivity, and Executive Function in Clinical Populations A Machine-Learning Approach to Modeling Cognitive Flexibility, Stress Reactivity, and Executive Function in Clinical Populations 1 11 10.61838/kman.jarac.879 EN Journal Article 2025 08 22 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> The objective of this study was to develop a machine-learning model integrating behavioral measures of executive function, cognitive flexibility, and psychophysiological stress reactivity to predict and classify clinical symptomatology in individuals with Major Depressive Disorder, Generalized Anxiety Disorder, and Post-Traumatic Stress Disorder.</p> <p><strong>Methods and Materials:</strong> A cross-sectional design was utilized with a sample of  participants, comprising Healthy Controls ( ) and individuals diagnosed with Major Depressive Disorder ( ), Generalized Anxiety Disorder ( ), and Post-Traumatic Stress Disorder ( ). Participants completed a comprehensive three-hour assessment protocol including neurocognitive tasks (Wisconsin Card Sorting Test, Stroop Color-Word Test, Trail Making Test), a modified social-evaluative stress task with concurrent psychophysiological recordings (electrocardiography for heart rate variability [HRV], salivary cortisol), and clinical self-report questionnaires. Machine learning models (Random Forest, Support Vector Machine, and XGBoost) were trained on of the data to classify clinical versus control groups, utilizing hyperparameter tuning and SHAP (SHapley Additive exPlanations) values for feature interpretation.</p> <p><strong>Findings:</strong> The XGBoost algorithm achieved the highest predictive performance in classifying clinical versus control groups, yielding an AUC-ROC of . Traditional linear analyses indicated that clinical cohorts exhibited significantly greater symptom severity and poorer cognitive performance, including elevated perseverative errors on the WCST and compromised Trail Making Test performance. Physiologically, cortisol reactivity ( ) was blunted in the PTSD group and exaggerated in the MDD group, while baseline HRV was significantly lower across all clinical cohorts. SHAP feature importance analysis revealed that the TMT Part B-A difference score was the most crucial predictor of clinical classification, followed closely by cortisol reactivity ( ), baseline anxiety, and baseline HRV.</p> <p><strong>Conclusion:</strong> Advanced machine-learning algorithms effectively capture the complex, non-linear interactions between cognitive flexibility and autonomic stress dysregulation, underscoring the necessity of integrated, multidimensional approaches for accurate psychiatric classification.</p> </td> </tr> </tbody> </table> https://journals.kmanpub.com/index.php/jarac/article/download/879/9412