<![CDATA[Student dropout remains a persistent challenge in State Islamic Higher Education because it is shaped by the interaction of academic performance, socio-economic conditions, psychological factors, and student engagement. Traditional approaches that rely only on static academic indicators such as Grade Point Average (GPA) often fail to capture early signals of dropout risk. This study aims to develop an early detection model for student dropout risk using the Support Vector Machine (SVM) algorithm by integrating academic, behavioral, socio-economic, and psychological indicators. This research employed a quantitative approach within an Educational Data Mining framework and followed the Knowledge Discovery in Databases (KDD) process. The dataset consisted of 467 undergraduate students with at least two semesters of academic records. Variables included GPA, GPA trend, attendance, Learning Management System (LMS) engagement, parental income, part-time work hours, financial support, motivation, and stress indicators. The results show that the optimized class-weighted SVM model using the Radial Basis Function (RBF) kernel achieved an accuracy of 81.9%, specificity of 88.1%, precision of 23.1%, and recall of 30.0% for the at-risk class. These findings indicate that the model is useful as an initial screening tool, especially for reducing false alarms among active students. However, its limited recall suggests that it should be combined with academic advisor verification and institutional follow-up. The study highlights that dropout risk is multidimensional and requires data-informed, timely, and human-supported intervention.]]>
