A Closed-Loop Artificial Intelligence System for Process-Oriented Student Assessment and Early Performance Prediction in Higher Education

The increasing amount of detailed learning data in higher education has shifted assessment methods from traditional scores to continuous assessment of student learning performance. This paper presents an applied closed-loop AI system that supports early performance prediction, informed risk decision-making, and personalized intervention planning. The system utilizes aggregated weekly data from learning management systems, including self-regulated learning metrics and participation indicators. A supervised regression model predicts near-future learning performance while employing chronological training strategies to prevent information leakage. Emphasizing decision-driven assessment, the system translates predictive outputs into a resource-sensitive decision framework to identify students requiring early support. Explainable artificial intelligence (XAI) techniques enhance the model’s transparency, providing insights into key learning factors and individual student-level explanations. The system generates actionable recommendations based on these insights. Its closed-loop architecture integrates data collection, prediction, interpretation, recommendation generation, and feedback logging, allowing instructors to monitor learning pathways and examine subsequent learning outcomes. Rather than claiming causal intervention effects, the proposed system illustrates how established AI, explainable learning analytics, and decision-support techniques can be operationally integrated to support ongoing student assessment in higher education.