An AI-Enabled, Data-Driven PDCA Framework for Continuous Quality Improvement in Higher Education: A Three-Institution Study

Abstract

Higher education institutions increasingly employ data analytics and artificial intelligence (AI) to enhance student learning outcomes, yet the absence of a systematic Continuous Quality Improvement (CQI) framework limits their effectiveness. This study aimed to develop and pilot an AI-enabled, data-driven Plan–Do–Check–Act (PDCA) framework to support CQI in higher education. Student-level learning records (n = 12,450) were preprocessed using z-score normalization and outlier removal at ±3 standard deviations, and analyzed through Random Forest and Support Vector Machine models. Predictive performance was evaluated using ten-fold cross-validation. Complementary qualitative data were collected from 18 faculty members and policy-makers via structured interviews to assess usability and impact on decision-making. The Random Forest model achieved 87.2% accuracy (area under the curve = 0.91), outperforming the Support Vector Machine model (82.5%, AUC = 0.86). Implementation of two PDCA cycles led to measurable improvements: course pass rates increased by 9.4%, and semester-on-time completion rose by 7.1% (Cohen’s d = 0.65, p < 0.01). Qualitative feedback indicated enhanced decision-making speed and higher stakeholder engagement. These results demonstrate that integrating predictive analytics with structured PDCA cycles can effectively support CQI, providing actionable insights to improve student success. The study contributes a practical, datadriven framework for higher education quality enhancement, highlighting the feasibility of AI-assisted monitoring and evaluation. While the pilot was limited to two semesters, the findings suggest that longitudinal application could sustain and further amplify improvements in institutional performance.