Personalized Learning Path Design for Vocational Education Based on CNN+Transformer Hybrid Architecture

Addressing the substantial individual differences among vocational education learners and the inadequate personalization of conventional learning pathways, this study introduces a novel personalized learning path design system leveraging a CNN+Transformer hybrid architecture. The proposed framework synergistically integrates the local pattern extraction capabilities of Convolutional Neural Networks (CNNs) with the global dependency modeling strengths of Transformers, establishing a comprehensive multi-task learning paradigm that simultaneously addresses performance prediction, learning path recommendation, and difficulty adaptation. The system employs a sophisticated dual-branch feature extraction mechanism, enhanced by a dense fusion strategy and joint optimization approach, to effectively capture and model the multi-dimensional characteristics of learning behaviors. Comprehensive experiments conducted on the xAPI educational dataset demonstrate the superior performance of our CNN+Transformer hybrid architecture, achieving 92.4% classification accuracy and 91.8% F1-score—representing substantial improvements of 3.3% and 3.5% over the best-performing baseline methods, respectively. Systematic ablation studies validate the contribution of each architectural component, with the feature fusion mechanism providing a 1.2% performance enhancement and the multi-task learning strategy contributing an additional 0.5% improvement. These findings establish our approach as an effective solution for advancing personalized educational technologies and demonstrate significant implications for the development of next-generation intelligent education systems.