APP下载

基于神经网络的非线性学术评价指标模拟权重研究

2021-04-02俞立平

现代情报 2021年4期

收稿日期:2020-08-01

基金项目:国家社会科学基金项目“学术评价与创新绩效评价问题研究”(项目编号:19FTQB011);浙江省一流学科A类项目(浙江工商大学统计学,管理科学与工程);浙江省自然科学基金重点项目“制造业从数量型创新向质量型创新转型机制研究”(项目编号:Z21G030004)。

作者简介:俞立平(1967-),男,教授,博士,博士生导师,研究方向:技术经济、科技评价。

摘要:[目的/意义]多属性评价方法在学术评价中应用广泛,其中以非线性评价方法为主,这些评价方法降低了权重的作用,使得评价指标与评价结果的关系不直观,也不利于对非线性评价方法进行评估和比较。[方法/过程]本文以因子分析为例,基于JCR2017经济学期刊数据,分别采用多元回归、岭回归、偏最小二乘法、BP神经网络计算模拟权重,以选取最优模拟权重估计方法。[结果/结论]在学术评价中有必要测度非线性评价的模拟权重;BP神经网络是计算模拟权重的最有效手段;模拟权重可以用来进行非线性学术评价方法的评估和选取;模拟权重的应用严重依赖数据。

关键词:学术评价;因子分析;模拟权重;BP神经网络

DOI:10.3969/j.issn.1008-0821.2021.04.013

〔中图分类号〕G302〔文献标识码〕A〔文章编号〕1008-0821(2021)04-0133-13

Research on Simulated Weights of Nonlinear Academic

Evaluation Index Based on Neural Network

Yu Liping

(School of Statistics and Mathematics,Zhejiang Gongshang University,Hangzhou 310018,China)

Abstract:[Purpose/Significance]Multi-attribute evaluation method is widely used in academic evaluation,among which the non-linear evaluation method is the main one.These evaluation methods reduce the role of weight,which makes the relationship between evaluation index and evaluation results not intuitive,and are not conducive to the evaluation and comparison of non-linear evaluation methods.[Method/Process]Taking factor analysis as an example,based on JCR2017 Economic Journal data,multiple regression,ridge regression,partial least square and BP neural network were used to calculate the simulated weights to select the optimal simulated weight estimation method.[Result/Conclusion]It is necessary to measure the simulated weight of nonlinear evaluation in academic evaluation;BP neural network is the most effective method to calculate the simulated weight;Simulated weight can be used to evaluate and select nonlinear academic evaluation methods;The application of simulated weight relies heavily on data.

Key words:academic evaluation;factor analysis;simulated weight;BP neural network

在学术评价中非线性评价方法是一种应用非常广泛的评价方法。从学术评价定量方法的角度,大致可以分为单指标评价与多属性评价两大类,如图1所示。学术评价单指标众多,典型的有h指数、影响因子、特征因子、扩散因子等。由于单指标评价所能提供的信息量有限,因此指标体系多属性评价方法就得到了广泛的应用。……

登录APP查看全文