摘要
文章以图书馆、情报与文献学CSSCI来源期刊为例,综合采用面板数据模型、面板变系数模型、面板门槛模型研究文章下载与被引之间的关系,并采用BP人工神经网络基于下载次数预测被引次数。研究表明:面板数据固定效应模型可有效分析下载次数与被引次数关系;论文被引次数主要受滞后2年的下载次数影响;不同期刊下载次数对被引次数的影响呈现趋同趋势;下载次数对被引次数的影响呈现非线性门槛特征;采用BP人工神经网络模型可以根据下载次数较好地预测被引次数;部分学科期刊可采用下载次数进行预评价,以提高评价的时效性。
Taking CSSCI journals of Library&Information Science as an example,this paper comprehensively adopts the panel data model,the panel variable coefficient model and the panel threshold model to study the relationship between downloads and citations,and uses the BP artificial neural network to predict the citations based on the number of downloads.The results show that:The panel data fixed-effects model can be used to effectively analyze the correlation between downloads and citations;The citation counts of papers are mainly affected by the number of downloads with a lag of 2 years;The influence of downloads on the citations in different journals shows a tendency of convergence with a nonlinear threshold characteristic;Adopting the BP artificial neural network model can better predict the number of citations based on downloads;The download counts of some journals in certain disciplines can be used to conduct pre-evaluations so as to improve the timeliness of the evaluation.
作者
俞立平
胡甲滨
YU Liping;HU Jiabin
出处
《图书馆论坛》
CSSCI
北大核心
2024年第5期77-85,共9页
Library Tribune
基金
国家社科基金后期资助项目“学术期刊评价--指标创新与方法研究”(项目编号:21FTQB016)研究成果。
