摘要
为更加精准地测量企业创新能力,利用熵值法和相关性分析,从创新投入、创新产出、创新支撑3个维度建立企业创新能力评价指标体系,构建企业创新能力评价的RBF-BP复合神经网络模型。该模型由1个输入层、1个RBF隐含层、1个BP隐含层以及1个输出层组成,其特点是将RBF隐含层的输出作为BP隐含层的输入。十折交叉验证与随机二次抽样2种方法检验表明,与单一RBF神经网络、单一BP神经网络相比,RBF-BP复合神经网络模型的平均均方误差与平均绝对误差分别下降28.21%、15.19%和12.51%、12.55%,表明RBF-BP复合神经网络模型具有最优的数据拟合能力,更适合于企业创新能力评价。
In order to have a more accurate measure of enterprise innovation ability,an evaluation index system is built from innovation input,innovation output,and innovation support using the entropy method and correlation analysis,and an RBF-BP composite neural network model made to evaluate enterprise innovation ability.The model consists of an input layer,an RBF hidden layer,a BP hidden layer,and an output layer,and is characteristic of the output of the RBF hidden layer as the input of the BP hidden layer.Compared with a single RBF neural network and a single BP neural network after the test of 10-fold cross-validation and random subsampling,the average mean square error and average absolute error of the RBF-BP composite neural network model decreased by 28.21%,15.19%and 12.51%,12.55%respectively,indicating that the RBF-BP composite neural network model has the best data fitting ability and is more suitable for the evaluation of enterprise innovation ability.
作者
张亚洲
李晓青
黄小博
ZHANG Yazhou;LI Xiaoqing;HUANG Xiaobo(School of Economics & Management,Xiamen University of Technology,Xiamen 361024,China)
出处
《厦门理工学院学报》
2021年第6期51-56,共6页
Journal of Xiamen University of Technology
基金
福建省社会科学基金项目(FJ2021B159)
福建省创新战略研究项目(科协联合2021R0160)
厦门理工学院研究生科技创新计划项目(YKJCX2020136,YKJCX2021044)。
关键词
企业创新能力
评价模型
RBF-BP复合神经网络
熵值法
相关性分析
enterprise innovation ability
evaluation model
RBF-BP composite neural network
entropy method
correlation analysis
