摘要
【目的/意义】针对学术期刊多指标评价中存在的复杂指标关系,提出一种新的基于深度自编码器的学术期刊综合评价方法。【方法/过程】考虑到评价指标间存在的复杂相关性,首先分析了两两指标之间的四种关系;其次,构造深度自编码器将多指标映射到三维隐元空间;然后,利用该空间分析学术期刊影响力;最后,利用非线性曲线拟合分析预测学术期刊的影响力指数。【结果/结论】以法学类期刊为研究对象,研究结果表明,该方法能够充分综合具有复杂关系的多指标因素,能够以可视化的方式展示学术期刊影响力,能够自动发现指标值异常的期刊,并能够应用学术期刊影响力分布规律精确预测学术期刊的影响力指数,是一种有效的学术期刊综合评价方法。
【Purpose/significance】Based on deep auto-encoder, this paper proposes a novel approach of comprehensive evaluation on academic journals by analyzing the complex relationship between multiple journal metrics.【Method/process】Considering such a complex relationship, the proposed approach first analyzes four kinds of relationships of pairwise metrics. Then, it constructs a deep auto-encoder to map multiple metrics of academic journals into a three-dimensional latent space, in which the impacts of academic journals are studied. Last, it performs a prediction of Academic Journal Clout Index(CI) by non-linear curve fitting.【Results/conclusions】The results on 94 law journals demonstrate that the proposed approach is an effective approach of evaluating the impacts of academic journals because it is able to integrate complex multiple metrics, visually exhibit the impacts of academic journals, automatically detect the journals having outlier properties,and provide a precise prediction of CI.
作者
徐小莹
李辉
XU Xiao—ying;LI Hui(Library of Northwestern Polytechnical University,Xi'an 710072,China)
出处
《情报科学》
CSSCI
北大核心
2019年第12期71-77,共7页
Information Science
基金
西北工业大学发展战略研究基金“双一流背景下高校图书馆智慧服务模式研究”(2019FZY14)
关键词
学术期刊
综合评价
深度学习
指标关系
academic journal
comprehensive evaluation
deep learning
index relationship
