基于BP人工神经网络的学术期刊组合评价

Combinatorial Evaluation of Academic Journals Based on BP Artificial Neural Network

在学术期刊评价中,组合评价得到了广泛应用,但组合评价方法和多属性评价方法均可用于组合评价问题,从而加剧了组合评价的复杂性,这种现象被称为组合评价悖论。为了解决这个问题,文章提出采用BP人工神经网络进行组合评价的思路,基于JCR2017经济学期刊,分别采用主成分分析、因子分析、TOPSIS进行单一评价,然后采用BP人工神经网络进行组合评价,并对评价效果进行分析。研究结果表明:组合评价悖论是学术评价中面临的新问题;BP人工神经网络是一种全新的组合方法,具有充分尊重单一多属性评价方法、精度高的优点,很大程度上解决了组合评价悖论问题;BP人工神经网络让评价重点重新回归到多属性评价,避免了评价方法过度技术化;BP人工神经网络组合评价不适合小样本。

In the evaluation of academic journals, combinatorial evaluation has been widely used.However, both the combination evaluation method and the multi-attribute evaluation method can be applied to the combinatorial evaluation problem, thus aggravating the complexity of the combinatorial evaluation, and this phenomenon is called the combinatorial evaluation paradox. In order to solve this problem, this paper proposes the idea of using BP artificial neural network for combinatorial evaluation. Based on JCR2017 economics journal, principal component analysis, factor analysis and TOPSIS are used for single evaluation, then BP artificial neural network is used for combinatorial evaluation, and finally the evaluation results are analyzed. The results indicate that the paradox of combinatorial evaluation is a new problem in academic evaluation, that BP artificial neural network is a new combinatorial method, which has the advantages of fully respecting single multi-attribute evaluation method and high accuracy,thus solving the paradox of combinatorial evaluation, that BP artificial neural network makes the evaluation focus return to multiattribute evaluation and avoids over-technicalization of evaluation methods, and that BP artificial neural network portfolio evaluation is not suitable for small samples.