CE ² ‐ GLF : A Knowledge‐Enhanced Graph Learning Framework for Expert Systems in Education‐to‐Employment Transition Modelling

Graph neural networks (GNNs) have shown great promise in modelling complex, heterogeneous and temporal relationships in domains such as education and employment. In this paper, we propose CE 2 ‐GL a cross‐domain graph learning framework that integrates temporal, heterogeneous and contrastive learning paradigms to model student learning trajectories and their transition to employment. The framework incorporates a heterogeneous relational encoder, a time‐aware adaptive evolution operator (T‐AEEO) and a course‐skill‐occupation alignment module (CSO‐Graph) to capture fine‐grained educational behaviours and labour market dynamics. Experiments on public data sets (EdNet, ASSISTments, O*NET) and a real‐world vocational graduate data set demonstrate that CE 2 ‐GLF not only improves prediction accuracy and representation quality but also provides interpretable insights for career guidance and educational decision support. The framework offers a scalable and principled solution for building knowledge‐driven expert systems in the education‐employment domain.