基于灰色神经网络的科技论文发表数预测研究

Research on Published Number Prediction of Scientific Papers Based on Grey Neural Network

[目的/意义]对我国科技论文发表数量进行较为准确的预测,旨在为有关部门制定相关政策提供参考。[方法/过程]选取我国2005—2019年的相关数据进行实证分析,运用GM-BP神经网络组合预测模型,将灰色系统理论中的灰色关联分析、GM(1,N)模型与BP神经网络相结合,提高了预测的准确度。首先,运用灰色关联分析法对相关因素进行评估,选取灰关联度较高的因素构建GM(1,N)模型;其次,运用GM(1,N)模型的预测结果连同相关因素数据训练BP神经网络;最后,用训练好的BP神经网络进行二次预测并输出为最终预测结果。[结果/结论]与传统GM(1,N)模型相比,所构建的GM-BP神经网络对于科技论文发表量的预测准确度更高。

[Purpose/significance]The accurate prediction of the number of scientific papers published in China aims to provide references for relevant departments to formulate policies.[Method/process]The relevant data of China from 2005 to 2019 are selected for empirical analysis and the combined forecasting model of GM-BP neural network is used.This method combines grey relational analysis in grey system theory GM(1,N)model and BP neural network and improves the forecasting accuracy.Firstly the grey relational analysis method is used to evaluate the related factors and the factors with higher grey relational degree are selected to construct the GM(1,N)model.Secondly BP neural network is trained by using the prediction results of GM(1,N)model and related factor data.Finally the trained BP neural network is used for secondary prediction and output as the final prediction result.[Result/conclusion]The results show that compared with the traditional GM(1,N)model the GM-BP neural network constructed in this paper has higher prediction accuracy for the published amount of scientific papers.