Industrial upgrading and regional coordination in the Guangdong-Hong Kong-Macao Greater Bay Area have accelerated, and the demand for cross-regional training, mobility and sharing of skilled personnel has continued to grow. In this paper, a computational research framework integrating multi-source heterogeneous data processing, knowledge graph modeling, collaborative network representation, spatio-temporal prediction and strategy recommendation is constructed for the joint construction and sharing scenario of skilled talents in the Bay Area, and a unified analysis of training collaboration, evaluation mutual recognition, employment and entrepreneurship, public services, industrial chain matching and open configuration is carried out. The experimental data covers January 2020 to December 2024, a total of 14,728 original samples are collected, and 13,860 are retained after cleaning. The results show that the comprehensive recognition scores of Shenzhen and Guangzhou are 0.912 and 0.901 respectively, the strength of Guangzhou-Shenzhen relationship is 0.91, the open configuration efficiency is improved from 0.62 to 0.86 under the comprehensive optimization scenario, the model prediction accuracy is 91.8%, RMSE and MAE are 0.052 and 0.039, respectively. The research shows that data fusion, graph relationship modeling and spatio-temporal prediction coupling can effectively improve the refined analysis ability of the co-construction and sharing research of skilled talents in the Bay Area, which is of practical significance for promoting regional collaborative governance and the construction of high-level skilled talents system.
