| 网络科技论文学术影响力评价指标的选择 |
| |
| 引用本文： | 沈小玲,严卫中.网络科技论文学术影响力评价指标的选择[J].图书情报工作,2013,57(3):69-77. |
| |
| 作者姓名： | 沈小玲 严卫中 |
| |
| 作者单位： | 1. 安徽财经大学图书馆;
2. Machine Learning Lab, GE Global Research Center Niskayuna |
| |
| 基金项目： | 国家社会科学基金一般项目"基于链接分析的网络科技论文学术影响评价研究" |
| |
| 摘 要： | 网络科技论文影响力的评价效果取决于评价指标变量的选择。将网络科技论文影响力评价与论文排名相关联,以Web of Science数据库中数学类论文为样本,从6个不同的排名等级组,即排名前0.01%、0.01%-0.1%、0.1%-1%、1%-10%、10%-20%、20%-50%,分别抽取论文数十篇,用文献信息方法对单篇论文从内容、论文载体和论文作者三个层面初选28个特征变量,以324篇论文的排名等级与28个学术链接指标样本建立为"序回归"模型的研究问题。基于Lasso方法对28个学术链接指标进行变量选择和参数估计,获得9个学术链接特征指标作为评价网络科技论文学术影响力的基本特征指标;以418篇OA论文的排名等级对23个网络影响计量及其衍生变量进行变量选择,获得5个论文网络传播与利用影响力的评价指标。最终共获得14个网络科技论文学术影响力的评价指标。
|
| 关 键 词： | 网络科技论文 特征指标 学术链接 影响力评价 Lasso回归 序回归 |
| 收稿时间： | 2012-09-28 |
Selection of Academic Influence Evaluation Index of Network Science and Technology Articles |
| |
| Shen Xiaoling,Yan Weizhong.Selection of Academic Influence Evaluation Index of Network Science and Technology Articles[J].Library and Information Service,2013,57(3):69-77. |
| |
| Authors: | Shen Xiaoling Yan Weizhong |
| |
| Institution: | 1. Library of Anhui University of Finance & Economics, Bengbu 233030;
2. Machine Learning Lab GE Global Research Center Niskayuna, New York 12065 |
| |
| Abstract: | The influence evaluation effect of network science and technology articles depends on the selection of evaluation index variable. Firstly, this paper associates influence evaluation of science and technology articles with their ranking in popular open-access journal databases. It takes mathematics articles from the WOS database as sample, and respectively selects ten articles from the top six ranking groups which are 0.01%、0.01%-0.1%、0.1%-1%、1%-10%、10%-20% and 20%-50%. With literature information method, 28 different features are extracted from each of these articles from three aspects of the article content, the journal publishing articles, and the authors. Then it takes the ranking level of 324 articles and 28 academic link indexes as the sample, to study the "ordinal regression model". Secondly, with Lasso method, based on the variable selection and parameter estimation of 28 academic link indexes, it obtains 9 features that are considered to be the basic characteristics indexes to evaluate academic influence of network science and technology articles. Thirdly, based on the ranking level of 418 OA articles, and the selection of 20 network influence measurement indexes and its derivative variable, this paper gets 5 evaluation indexes of network transmission and utilization of impact. Finally, this paper generalizes 14 academic influence evaluation indexes of network science and technology articles. |
| |
| Keywords: | feature selection academic hyperlink influence assessment lasso regression ordinal regression |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《图书情报工作》浏览原始摘要信息 |
| 点击此处可从《图书情报工作》下载免费的PDF全文 |
|
