A KG-DRL framework for post course competition certificate integration and path generation in civil aviation vocational colleges

Abstract

Abstract The “Post-Course-Competition-Certificate” (PCCC) integration model has emerged as a cornerstone of vocational education reform in China, yet its implementation in civil aviation vocational colleges is hindered by the absence of quantifiable fusion criteria, reliance on subjective evaluation practices, and the lack of adaptive learning pathways tailored to individual student profiles. This paper proposes a unified framework that jointly addresses these limitations through knowledge graph (KG) reasoning and deep reinforcement learning (DRL). We construct a heterogeneous KG that semantically integrates job posts, course syllabi, skill competitions, and vocational certificates, and derive a composite integration degree metric comprising semantic matching, structural association, and path reachability. The personalised learning path generation problem is subsequently formulated as a Markov decision process (MDP) and solved via a hybrid Deep Q-Network (DQN) and Proximal Policy Optimisation (PPO) algorithm, where the reward function explicitly promotes fusion gain while respecting curriculum and regulatory constraints. Experimental results on data from a civil aviation vocational college demonstrate that the proposed framework achieves substantial improvements over traditional methods across all evaluation metrics, including global integration degree, path efficiency, and job matching similarity, while generating admissible curricula that satisfy mandatory credit and regulatory requirements. The results validate the effectiveness of the framework in delivering quantifiable, personalised, and regulatable learning pathways for vocational education.