Predictive Analytics In University Erp Systems: Enhancing Student Success Through Artificial Intelligence

Abstract

University enterprise resource planning (ERP) systems contain longitudinal academic, administrative, financial, 
and engagement traces, yet these traces are commonly used only for retrospective reporting. This empirical paper 
examines whether an artificial-intelligence (AI) layer embedded in a university ERP can identify students at risk 
early enough to improve student success. A quasi-experimental study was designed around 2,544 undergraduate 
students enrolled across six programmes at a large public university. Historical records (n = 1,260) formed the 
comparison cohort, while a subsequent cohort (n = 1,284) received weekly risk scoring and advisor-facing 
explanations. De-identified ERP, learning-management-system, attendance, and support-service data were 
transformed into 34 temporal and cumulative features. Gradient boosting was selected after stratified validation; 
it achieved an area under the receiver-operating-characteristic curve of 0.87 and recalled 76% of eventual non
completers at the operating threshold. Alerts were not sent to students automatically: trained advisors reviewed 
evidence, contacted students, and documented referrals. Compared with the historical cohort, the AI-supported 
cohort showed higher course completion (84.9% versus 78.4%), term retention (89.3% versus 84.1%), and mean 
grade-point average (7.18 versus 6.82 on a 10-point scale). Gains were largest among students with low 
attendance and delayed fee clearance, but false positives and uneven intervention capacity remain material 
concerns. The findings support an augmentation model in which ERP analytics prioritizes human care rather than 
automates high-stakes decisions. The paper contributes a transparent measurement framework, implementation 
evidence, and governance conditions for responsible predictive analytics in higher education.