A dynamic planning method for personalized growth paths of secondary vocational school students optimized by reinforcement learning

The growing diversity among learners in vocational education necessitates the development of intelligent, adaptive systems that cater to individual differences in academic performance, engagement, and skill development. Conventional planning approaches tend to neglect these variations, resulting in inefficiencies in personalized learning. This research announces a dynamic reinforcement learning approach, Snap-Drift Cuckoo Search-based Deep Deterministic Policy Gradient (SBCS-DDPG), that optimizes personalized growth paths for secondary vocational students. The SBCS-DDPG model unites the exploratory optimization ability of SBCS with the adaptive decision-making capability of DDPG, enabling continuous refinement of individualized learning trajectories. The SBCS algorithm employs Lévy flight-based exploration and adaptive drift–snap dynamics to optimize DDPG’s hyperparameters, such as learning rate, exploration noise, and update coefficients. This ensures efficient convergence, robust exploration, and stability in learning performance. The DDPG component then operates within the reinforcement learning framework, where each student’s academic, behavioral, and skill profile represents a state, and recommended learning modules or interventions represent actions. The model’s reward function maximizes long-term outcomes in academic achievement, skills acquisition, and engagement, while a penalty mechanism minimizes stagnation and inefficiency. The model was trained and tested using data from 658 vocational students, with an 80:20 training-testing split. Experimental evaluation on a personalized learning path recommendation task demonstrates that the proposed SBCS-DDPG framework achieves 97.8% accuracy, 98.2% precision, and an F1-score of 98%, outperforming established baseline models. Additionally, Shapley Additive Explanations (SHAP) were utilized in the research. This confirms the SBCS-DDPG model’s capacity for adaptive, scalable, and data-driven personalization in vocational education.