Scholarlink

<p>In response to the problems existing in the evaluation of scientific research talents in China, such as insufficient scientificity of evaluation standards, strong subjectivity in the evaluation process, poor accuracy of the results, and the weak operability of traditional evaluation indicators after refinement, which makes it difficult to comprehensively reflect the comprehensive strength of talents, this paper aims to construct a scientific, comprehensive and feasible evaluation index system and an accurate evaluation prediction model for scientific research talents based on the research results of the talent profiling system and the concept of comprehensive competence elements. This will provide guidance for the practice of evaluating scientific research talents. Firstly, through questionnaire surveys combined with the opinions of experts in the field of talent evaluation, the initial evaluation index system was constructed using the Delphi method and the entropy method. After correlation analysis and expert verification, the indicators were selected and the weights of each indicator were determined. Then, a 4-layer fully connected BP neural network was constructed, and comparative experiments were designed with input layers containing 6, 10 and 21 evaluation indicators. Using 47 sets of questionnaire survey data as training samples and 23 sets of scraped data as test samples, the model performance was verified. Finally, an evaluation system consisting of 5 first-level indicators (research management, basic qualities, research contribution, research potential, and academic status), 10&#xa0;s-level indicators, and 21 third-level indicators was established. The comparative experiments showed that the BP neural network model with 6 most correlated indicators in the input layer had the smallest prediction error, with an average mean square error of 0.0367 and the best generalization ability. This model was applied to the evaluation of 23 professional scientific research talents at BJ University, successfully achieving the precise prediction of the comprehensive strength of talents in different fields. The evaluation index system for talents constructed in this study comprehensively covers the core competence elements of scientific research talents, and the evaluation prediction model based on the BP neural network has high accuracy and practicality.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Construction of Talent Evaluation Index System and Evaluation Prediction Model Based on BP Neural Network

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Qiong Cao,
  • Lizhi Pan,
  • Jie Zhao,
  • Zhonghao Liang,
  • Muqing Wang

摘要

In response to the problems existing in the evaluation of scientific research talents in China, such as insufficient scientificity of evaluation standards, strong subjectivity in the evaluation process, poor accuracy of the results, and the weak operability of traditional evaluation indicators after refinement, which makes it difficult to comprehensively reflect the comprehensive strength of talents, this paper aims to construct a scientific, comprehensive and feasible evaluation index system and an accurate evaluation prediction model for scientific research talents based on the research results of the talent profiling system and the concept of comprehensive competence elements. This will provide guidance for the practice of evaluating scientific research talents. Firstly, through questionnaire surveys combined with the opinions of experts in the field of talent evaluation, the initial evaluation index system was constructed using the Delphi method and the entropy method. After correlation analysis and expert verification, the indicators were selected and the weights of each indicator were determined. Then, a 4-layer fully connected BP neural network was constructed, and comparative experiments were designed with input layers containing 6, 10 and 21 evaluation indicators. Using 47 sets of questionnaire survey data as training samples and 23 sets of scraped data as test samples, the model performance was verified. Finally, an evaluation system consisting of 5 first-level indicators (research management, basic qualities, research contribution, research potential, and academic status), 10 s-level indicators, and 21 third-level indicators was established. The comparative experiments showed that the BP neural network model with 6 most correlated indicators in the input layer had the smallest prediction error, with an average mean square error of 0.0367 and the best generalization ability. This model was applied to the evaluation of 23 professional scientific research talents at BJ University, successfully achieving the precise prediction of the comprehensive strength of talents in different fields. The evaluation index system for talents constructed in this study comprehensively covers the core competence elements of scientific research talents, and the evaluation prediction model based on the BP neural network has high accuracy and practicality.