基于SOFM人工神经网络的长江三角洲地区城市职能分类

FUNCTION CLASSIFICATION OF SEVERAL CITIES IN THE YANGTZE DELTA BASED ON SOFM NEURAL NETWORK

在研究文献的基础上,提出了以人工神经网络A(NN)作为城市职能聚类的模型工具。结合实际,建立了包含城市规模、职能活动和产业活动的分类指标体系,并应用ANN模型分析中自组织特征映射网络S(elf-Orga-nizaing Feature Maps,SOFM)对长江三角洲地区的主要城市进行了职能分类。网络运行结果表明:2001年长江三角洲地区的15个主要城市可以自动分为5类,SO FM分类结果与实际情况基本一致;利用系统聚类和主成分分析方法对人工神经网络自动聚类的结果进行了验证,研究说明了SOFM神经网络方法更可行。

The method of function classification of cities by artificial neural networks has been put forward with literatures reviewed.According to the local situation, the index system including cities size, function activities and industry activities has been set up,and function classification of several cities in the Yangtze Delta done by a soundly trained self-organizing feature maps （SOFM）.The run results of SOFM show that 15 main cities can be classified into 5 groups, which are in accord with the local situation.As a result,it is revealed the method is more feasible than systematical c, lassification method and principal component method in the light of the classified results.