Is it Feasible to Use BP Neural Network for Academic Journal Evaluation: an Attempt Based on 6 Single Objective Weighting Methods

Abstract

In the new era, the analysis of academic journal evaluation methods and the comprehensive comparison of the advantages, disadvantages, and stability of different methods can help to provide some reference for academic journal evaluation. In this study, single model evaluation was carried out for 6 weighting methods without comprehensive evaluation value, and fuzzy comprehensive evaluation is performed on the results passing the nonparametric test. Based on the evaluation, BP neural network is introduced, and BP neural network evaluation model is established. The results show that the fuzzy Borda evaluation can integrate the evaluation value and evaluation order of single models, and has higher accuracy compared with single evaluation models. The prediction rate of the network model based on the gradient descent optimization algorithm can reach more than 80%, and the weights obtained from the continuous self‐learning of the neural network training set can reduce the subjectivity and mutual interference between indicators.