摘要： 针对神经网络建模需尽量减少输入变量及输入变量要充分解释输出变量的问题，研究了变量聚类分析方法，提出了变量选取原则并对这些原则进行了基于信息论的理论推导，开发了基于聚类分析的神经网络软件。实际应用表明，这种基于精确数学分析进行的神经网络变量的选取是行之有效的。

关键词: 变量聚类, 神经网络, 信息论

Abstract: Aiming at the problems of building the model of neural network that the input variables should be reduced as less as possible and explain the output variables fully, the method of variable cluster analysis is researched. Some principles of selecting variables are given and deduced on the theory of information. The software of neural network based on cluster analysis is exploited. In the application of practice, the way of selecting the variables of neural network based on precise mathematical analysis is verified to be feasible and effective.

中图分类号: 

• TP183