A Random Forest-Based Early Warning System for Early Identification and Classification of Student Academic Risk Using Academic Performance Indicators

Abstract

Student attrition remains a major challenge in higher education, influencing institutional performance, student retention, and academic quality. This study proposes a Random Forest-based Early Warning System (EWS) integrated into the Global Academic Information System (GAIS) at Global Institute of Technology & Business to identify students at risk of academic failure. This model was developed using data from 1,485 students collected between 2020 and 2025, utilizing eight academic indicators: study program, Semester Grade Point Average (SGPA), cohort year, Cumulative Grade Point Average (CGPA), attendance rate, academic progress, total Semester Credit Units (SKS) completed, and repeated courses. Administrative aspects were excluded from this study to maintain the analysis's initial focus on academic indicators. Subsequently, students were classified into three risk categories: Active, Warning, and At-Risk of Dropout. Test results revealed significant characteristic differences between the groups; students in the Active category demonstrated superior CGPA, attendance rates, and academic progress compared to those in the dropout-risk category. The proposed model effectively identifies vulnerable students and provides timely risk classification to support academic intervention. By integrating predictive analytics into GAIS, the proposed EWS enables proactive monitoring and supports data-driven decision-making for improving student retention and institutional academic management