基于k-means聚类分析的高校论文统计研究

Statistical Research of University Papers Based on K-means Cluster Analysis

科研论文的质量能反映学者、学术机构和学术团队的科研水平。文章选取浙江理工大学、浙江工业大学、浙江师范大学等十所高校,提取各高校在2012—2016年所有被WOS核心合集中SCI数据库收录的论文,采用非监督的机器学习k-means算法对发文量、去自引被引频次、去自引施引文献及发文量权值等四个特征变量进行统计分析。结果表明:在这十所高校中,浙江工业大学和宁波大学属于第一等级,浙江理工大学、浙江师范大学和杭州师范大学属于第二等级,中国计量大学、杭州电子科技大学、温州大学、浙江农林大学和浙江工商大学属于第三等级。文章研究表明,利用k-means方法横向比较高校科研论文质量具有可行性。

The quality of research papers can reflect scientific research level of scholars,academic institutions and academic team.Ten universities such as Zhejiang Sci-Tech University,Zhejiang University of Technology and Zhejiang Normal University were chosen,and their articles＇ data downloaded from SCI database of ＂core collection＂in WOS（Web of Science,WOS）during five years（2012—2016）were extracted in this paper.K-means,an unsupervised algorithm,was employed for statistical analysis of four characteristic variables including quantity of publications,citation frequency without self-citation,citing articles without self-citation and weight of publications.The results showed that among these ten universities,Zhejiang University of Technology and Ningbo University are clustered to the first level;Zhejiang Sci-Tech University,Zhejiang Normal University and Hangzhou Normal University fall into the2 nd level and the other five universities（China Jiliang University,Hangzhou Dianzi University,Wenzhou University,Zhejiang A F University and Zhejiang Gongshang University）belong to the 3rd level.The study showed that it is feasible to apply k-means for horizontal comparison of the quality of universality papers.