融合BP神经网络的学术话语权评价方法探讨

Research on the Evaluation Method of Academic Discourse Power by Fusing on BP Neural Network

[目的/意义]学术话语权是中国国际话语权体系中不可分割的一部分,是国家政治、经济、科学技术软实力的主要表现,对学术话语权评价方法进行剖析,综合比较不同方法的优劣与稳定性,有助于为学术话语权评价提供一定参考。[方法/过程]采用6种无需综合评价值的赋权法进行单一模型评价,对通过非参数检验后的结果进行模糊组合评价,减少单一评价倾向,提高评价公信力,并引入BP神经网络,基于梯度下降算法构建神经网络模型。[结果/结论]构建基于创新引领指数的学术引领力、基于引文分析指标的学术影响力与基于Altmetrics指标的学术传播力三维度的学术话语权评价体系,模糊Borda评价可以综合考虑单一模型的评价值大小与评价序相对秩次,实现客观信息的内部组合,相较于单一评价模型有更高的准确度,并基于此构建了融合BP神经网络的学术话语权评价模型。

[Purpose/Significance]Academic discourse power is an integral part of China’s international discourse power system and the main manifestation of national political,economic,scientific and technological soft power.Analyzing the evaluation methods of academic discourse power and comprehensively comparing the advantages,disadvantages and stability of different methods are helpful to provide certain references for the evaluation of academic discourse power.[Method/Process]Six assignment methods without comprehensive evaluation values were used for single model evaluation,and fuzzy combination evaluation was carried out on the results after passing the non-parametric test to reduce the tendency of single evaluation and improve the credibility of evaluation,and BP neural network was introduced to construct neural network model based on gradient descent algorithm.[Result/Conclusion]This paper constructs an academic discourse power evaluation system based on three dimensions:academic leadership based on the innovation leading index,academic influence based on the citation analysis index and academic communication capacity based on the Altmetrics index.Fuzzy Borda evaluation allows for a combination of high and low evaluation scores and evaluation ranking for a single model,and realizes the internal combination of objective information,which has higher accuracy than the single evaluation model.Based on this,an evaluation model of academic discourse power integrated with BP neural network was constructed.