Coupling of High-Skilled Talent Cultivation in the Manufacturing Industry and Regional Economic Performance in Vocational Education by Integrating Machine Learning and DEA Models

This study develops a two-stage analytical framework integrating super-efficiency Slacks-Based Measure Data Envelopment Analysis (SBM-DEA), extreme gradient boosting (XGBoost), and SHapley Additive exPlanations (SHAP) to investigate the coupling relationship between high-skilled talent cultivation in vocational education and regional economic performance in manufacturing industries. The super-efficiency SBM-DEA model is employed to quantify regional talent–economy coupling efficiency, while XGBoost captures nonlinear interactions among educational, industrial, and innovation-related factors. SHAP-based interpretability analysis is further conducted to identify the dominant drivers and their marginal effects. Experimental results show that the proposed framework achieves a prediction goodness-of-fit (R2) of 0.91 and reveals the industrial upgrading index and dual-qualified teacher ratio as the most influential factors, with mean absolute SHAP values of 0.128 and 0.095, respectively. Dependency analysis further identifies a critical industrial upgrading threshold near 2.8, beyond which marginal benefits gradually diminish. The findings highlight the importance of coordinated talent development in supporting intelligent manufacturing, industrial communication systems, and digital production infrastructures, where the integration of skilled technical personnel, information technologies, and industrial innovation is essential for sustainable regional development. This study provides a data-driven engineering methodology for evaluating and optimizing the collaborative evolution of vocational education and manufacturing economies.