Data-Driven Evaluation of Institutional Performance Using AI Models
Year: 2026
Type: conference-paper
Abstract: As needs to enhance educational performance and operational effectiveness have grown at an alarming rate, there has been not much that higher education institutions can do other than react to retrospective data analysis. The study presents a multi-faceted, AI-based concept of institutional performance measurement, incorporating multi-faceted data streams, including academic scores, engagement metrics and resource consumption. We compare three architectural approaches: conventional Machine Learning (Random Forest), Ensemble Learning (XGBoost) and Deep Learning (Artificial Neural Networks). Through experiments on a simulated dataset of 1,000 student records, we show that complex models deliver superior performance over simple models. In particular, both XGBoost and Deep Learning models reached a maximum predictive accuracy of 97.5% (1.5% higher than Random Forest). Our study reveals that although Deep Learning excels in pattern analysis, XGBoost offers a more efficient and interpretable model for institutional governance. Our framework offers a dependable early detection system to identify vulnerable students, and a guide for policy-making in contemporary educational systems.
Institutions: Jesselton University College, Kingdom University
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: hybrid
OpenAlex ID: W7214775294