<![CDATA[Academic burnout among university students is an increasingly pressing issue due to its impact on learning motivation, attendance, and academic performance. Meanwhile, conventional early warning approaches tend to be subjective and lack systematic, data-driven prioritization mechanisms. This study develops a decision support system (DSS) integrating Extreme Gradient Boosting (XGBoost), Decision Making Trial and Evaluation Laboratory (DEMATEL), and Measurement of Alternatives and Ranking according to Compromise Solution (MARCOS) to prioritize academic burnout interventions. XGBoost was applied to a dataset of 60 studentscomprising eleven academic, behavioral, and psychosocial attributesto identify burnout indicators through feature importance analysis. Eight selected indicators were then evaluated by academic experts using DEMATEL to model cause-effect relationships and determine criterion weights based on inter-criterion dependencies. These weights were utilized in MARCOS to rank students according to intervention urgency. The XGBoost model achieved a five-fold cross-validation accuracy of 70.0% (SD = 13.5%), despite the limited sample size. DEMATEL identified Academic Stress and Sleep Duration as the most influential "effect" criteria, while Financial Stress and Study Load emerged as the dominant "cause" criteria. The MARCOS rankings showed strong correlations with Burnout Risk Scores (Spearman’s rho = 0.766; p < 0.001) and Intervention Urgency (rho = 0.714; p < 0.001). Students with high burnout levels consistently occupied the highest priority positions, with an average rank of 5.6 out of 60 students. These results demonstrate that integrating machine learning with DEMATEL–MARCOS yields an objective, explainable, and practical DSS framework to support interventions for student academic burnout.]]>
Copyrights © 2026
