基于因子-BP人工神经网络的期刊评价方法选择研究

目的/意义] 学术评价中多属性评价方法有几十种，每种评价方法结果均不相同，难以进行方法选择。方法/过程] 本文在分析评价本质属性的基础上，提出了一种新的多属性评价方法选取方法：因子-BP人工神经网络筛选法，并以JCR2017经济学期刊为评价对象，同时采用专家会议赋权法、主成分分析、因子分析、TOPSIS进行评价，并进行评价方法的选择。结果/结论] 研究结果表明：在学术评价中多属性评价方法的选择必须兼顾主观和管理因素；因子-BP人工神经网络筛选法是一种有效的评价方法选取手段；因子-BP人工神经网络筛选法可以用来辅助专家赋权；对于多属性评价方法的进一步优化成为新的问题；当评价对象较少时不宜采用本文方法。

Research on the Selection of Academic Evaluation Methods Based on Factor-artificial Neural Network-Taking Multi-attribute Evaluation Method as an Example

Purpose/Significance] There are dozens of multi-attribute evaluation methods in academic evaluation,and the results of each method are different,thus,it is difficult to choose a method.Method/Process] Based on the analysis of the essential attributes of evaluation,this paper proposed a new way of selecting multi-attribute evaluation methods:factor-artificial neural network screening.This paper took JCR2017 economic journal as the evaluation object while adapting the expert meeting weighting method,principal component analysis,factor analysis and TOPSIS to evaluate and select the evaluation method.Result/Conclusion] The results showed that the choice of multi-attribute evaluation method in academic evaluation must consider both subjective and management factors;factor-artificial neural network screening was an effective method to select evaluation methods;factor-artificial neural network screening can be used to assist expert empowerment;further optimization of multi-attribute evaluation method becomes a new problem;this method should not be used when there are few evaluation objects.