Scholarlink

<p>In the evolving landscape of scholarly evaluation, there is a pressing need for interpretable and data-driven methods to identify high-impact researchers. While platforms such as Google Scholar and Web of Science provide extensive bibliometric metadata such as total publications, citation counts, and h-index values, there remains no universally accepted framework to recognize top-performing academics. This study proposes a structured approach for learning interpretable decision rules from large-scale bibliometric data, aimed at identifying award-worthy researchers. Using a curated dataset of 1180 researchers in the civil engineering domain (590 awardees and 590 non-awardees), we compute 64 quantitative author assessment parameters spanning four categories. Feature importance is ranked using a multilayer perceptron (MLP) with recursive feature elimination, and decision trees are then used to derive transparent rules for researcher recognition. Our framework achieves classification accuracies between 60 and 69% and demonstrates that top-ranked parameters from each category effectively position 25–61% of awardees among the top 100 researchers. The findings offer a scalable and interpretable model for academic impact assessment and contribute to the development of objective recognition systems within the research community.</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

Knowledge discovery through interpretable decision rules: a framework for ranking researchers from bibliometric data

  • 期刊论文

  • 发表日期：

    2025年9月27日

查看全文
  • Ghulam Mustafa,
  • Muhammad Tanvir Afzal,
  • Yasir Noman Khalid

摘要

In the evolving landscape of scholarly evaluation, there is a pressing need for interpretable and data-driven methods to identify high-impact researchers. While platforms such as Google Scholar and Web of Science provide extensive bibliometric metadata such as total publications, citation counts, and h-index values, there remains no universally accepted framework to recognize top-performing academics. This study proposes a structured approach for learning interpretable decision rules from large-scale bibliometric data, aimed at identifying award-worthy researchers. Using a curated dataset of 1180 researchers in the civil engineering domain (590 awardees and 590 non-awardees), we compute 64 quantitative author assessment parameters spanning four categories. Feature importance is ranked using a multilayer perceptron (MLP) with recursive feature elimination, and decision trees are then used to derive transparent rules for researcher recognition. Our framework achieves classification accuracies between 60 and 69% and demonstrates that top-ranked parameters from each category effectively position 25–61% of awardees among the top 100 researchers. The findings offer a scalable and interpretable model for academic impact assessment and contribute to the development of objective recognition systems within the research community.