基于可解释机器学习的科技人才流动宏观影响因素研究OA

To accurately identify key influencing factors and their impact mechanisms on science and technology talent mobility,this study employed panel data of China's 31 provinces(municipality,autonomous region)from 1978 to 2023 to construct an XGBoost-based prediction model for science and technology talent mobility.The SHAP interpretation method was then introduced to conduct global interpretation of feature importance,feature dependency analysis,feature interaction analysis,and regional interpretation.The results indicate that:science and technology talent mobility is most significantly influenced by the research environment,followed by the education environment,economic development level,income level,and living convenience,while the impact of industrial structure is relatively weak;The research environment exhibits a positive linear relationship with science and technology talent mobility,whereas other factors demonstrate complex nonlinear relationships;Significant synergistic effects exist between economic development level and research environment,as well as between research environment and living convenience;Significant regional variations exist in the dominant drivers of talent mobility across eastern,central,western,and northeastern China.Based on these findings,policy recommendations are proposed to promote the rational mobility of science and technology talents among regions.