| 吴冰,齐思贤.集成传统学术评价和Altmetrics指标的论文高被引预测研究[J].数字图书馆论坛,2023,(9):30~37 |
| 集成传统学术评价和Altmetrics指标的论文高被引预测研究 |
| Research on High Citation Prediction of Papers by Integrating Traditional Academic Evaluation and Altmetrics Indicators |
| 投稿时间：2023-06-28 |
| DOI：10.3772/j.issn.1673-2286.2023.09.004 |
| 中文关键词: 论文引用；高被引预测；替代计量学；引用理论；机器学习 |
| 英文关键词: Paper Citation; High Citation Prediction; Altmetrics; Citation Theory; Machine Learning |
| 基金项目: |
| 作者 | 单位 | | 吴冰 | 同济大学经济与管理学院 | | 齐思贤 | 同济大学经济与管理学院 |
|
| 摘要点击次数: 776 |
| 全文下载次数: 583 |
| 中文摘要: |
| 随着Web 2.0和社交网络的发展，补充学术成果评价的Altmetrics指标应运而生，已有研究表明Altmetrics指标与被引频次之间存在相关性，但集成Altmetrics指标的论文高被引预测研究较少。因此，基于引用理论，将Altemetrics指标与学术层面指标相结合，构建论文高被引预测的指标体系；选取ESI高被引论文榜单，获取2022年4月经济与商业学科高被引论文合集，由此从Web of Science数据库获取论文集相关的学术层面数据，并从Altmetric LLP平台获取论文集相关的Altmetrics指标数据；经过数据清洗和预处理，共得到27 953篇论文数据，对比3种常用机器学习算法的论文高被引预测结果，得到最优的预测模型。研究结果表明：相较于仅使用学术层面指标，引入Altmetrics指标的论文高被引预测效果更优；Altmetrics指标中的在线阅读平台读者数对论文被引频次的影响最大，随后是学术层面指标中的期刊被引半衰期、论文首次被引两年内被引频次、一作总被引频次。研究可以为探究论文高被引的影响因素及其影响程度，完善学术成果的评价体系提供理论依据。 |
| 英文摘要: |
| With the development of Web 2.0 and social networks, Altmetrics indicators have emerged as supplementary evaluation of academic achievements. Previous studies have shown that there is a certain correlation between Altmetrics indicators and citation frequency, but there is limited research on high citation prediction of papers that integrates Altmetrics indicators. Therefore, based on the citation theory, this study combines Altmetrics indicators with academic indicators to construct an indicator system for predicting high citation in papers. Then, the ESI highly cited paper list is selected to obtain highly cited paper collection of April 2022 in economic and business discipline. Thereby, academic data relating with the paper collection are obtained from the Web of Science database, and Altmetrics indicator data relating with the paper collection is obtained from the Altmetric LLP platform. After data cleaning and preprocessing, data of 27 953 papers are obtained, and the evaluation results of three common m chine learning algorithms are compared to get the optimal model. The research results indicate that compared with using academic indicators alone, the integration of Altmetrics indicators yields better prediction of highly cited papers. The number of readers on online reading platforms has the greatest impact on the citation frequency of papers among Altmetrics indicators, followed by the journal’s citation half-life, the number of citations within the first two years after a paper’s initial citation, and the total number of citations as the first author among academic indicators. This study can contribute to exploring the factors influencing high citation and their respective impact levels, providing a theoretical basis for improving the evaluation system of academic achievements. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
