Knowledge-Based Recommendation for Graduate Subject Allocation Using Graph Attention Networks (GAT)
This study proposes a hybrid artificial intelligence framework for graduate subject allocation that enhances fairness, transparency, and operational efficiency in higher education. By integrating rule-based reasoning with advanced deep learning models, the framework supports data-driven decision-making for academic suitability, workload equity, and research alignment, while embedding Explainable Artificial Intelligence (XAI) to facilitate digital transformation in Thai universities. Traditional subject allocation processes in graduate programs are often manual, time-consuming, and subject to subjective judgment, thereby limiting their capacity to adapt to increasing curricular complexity and diverse faculty profiles. To address these challenges, the proposed framework combines institutional knowledge-based rules with machine learning techniques to model complex academic relationships across faculty, subjects, and workloads. Empirical analysis was conducted using data from 480 faculty members at Mahasarakham University and evaluated using multiple predictive models, including XGBoost, Wide-and-Deep Neural Networks, and Graph Neural Networks. Faculty performance was assessed using 20 institutional indicators reflecting teaching experience, research productivity, supervision, administrative responsibilities, and digital pedagogical competence. A multi-objective ranking algorithm was applied to simultaneously optimize academic suitability, workload balance, and research–teaching alignment. To enhance transparency and usability, the system incorporates a Faculty Subject Allocation Dashboard (FSAD) with SHAP-based interpretability, enabling administrators to understand and validate allocation decisions in real time. The results demonstrate that the proposed framework significantly improves allocation accuracy, workload equity, and decision transparency, offering a scalable and explainable solution for AI-driven academic governance.
