地理研究 ›› 2014, Vol. 33 ›› Issue (9): 1736-1746.doi: 10.11821/dlyj201409014

• • 上一篇    下一篇

基于PCA-SOM的深圳产业空间结构研究

章文1,2(), 王佳璆3   

  1. 1.中山大学地理科学与规划学院,广州510275
    2.广东省城市化与地理环境空间模拟重点实验室,广州510275
    3. 伦敦大学学院高级空间分析中心,伦敦 W1T 4TJ
  • 收稿日期:2013-06-11 修回日期:2013-10-23 出版日期:2014-09-20 发布日期:2014-11-10
  • 作者简介:

    作者简介:章文(1978-),男,安徽安庆人,博士生。主要从事城市产业规划与模拟研究。E-mail:zhangwen@ustc.edu.cn

  • 基金资助:
    国家自然科学基金(41001385);国家科技支撑计划项目(2012BAJ15B02)

A research on Shenzhen’s industrial spatial structure based on PCA-SOM

Wen ZHANG1,2(), Jiaqiu WANG3   

  1. 1. Geography and Planning School of Sun Yat-sen University, Guangzhou 510275, China
    2. Key Laboratory of Urbanization and Geographical Environment Simulation of Guangdong Province, Guangzhou 510275, China
    3. Centre for Advanced Spatial Analysis of University College London, London W1T 4TJ, Britain
  • Received:2013-06-11 Revised:2013-10-23 Online:2014-09-20 Published:2014-11-10

摘要:

利用深圳市企业空间分布数据,以街道为基本空间单元,运用主成分分析提取因子达到产业变量降维和抽象目的,在此基础上构建SOM神经网络进行聚类分析,通过PCA-SOM耦合模型实现了对城市内部产业空间分类和结构描述。研究表明:街道的产业功能可以通过提取各产业企业分布数据的主成分因子来表征;PCA-SOM耦合模型将深圳产业空间划分为6种类型区,分类结果与实际吻合;深圳城市产业结构存在空间分异,有别于传统城市的同心环形模式,深圳城市中心区位于地理空间的底部,以集聚型现代服务业为中心、整体呈扇型辐射,并具有明显的路径依赖特征;深圳东部分区产业功能现状仍不明朗,需要对该区域做进一步的空间结构规划和政策支持,以形成多中心的产业辐射模式。

关键词: 城市结构, 产业功能, 主成分分析, SOM聚类, 深圳

Abstract:

By using spatial distribution data of Shenzhen enterprises, regarding subdistricts as the basic spatial unit, this paper adopts the principal component analysis (PCA) method to extract factors to achieve the objective of dimension reduction and properties abstraction on industry variables. Based on the above, self-organizing map (SOM) neural network clustering model is built with three industrial characters of the subdistricts resulted from PCA and factors extraction as input data. Through PCA-SOM coupling model, this paper brings about the spatial classification and description of urban internal industrial function structure. Then the following conclusions are drawn. (1) The industrial functions of subdistricts can be represented through extracting the factors of enterprises spatial distribution data by principal component analysis. (2) PCA-SOM coupling model divides Shenzhen’s industrial spatial function into 6 type-zones. In this way the classification results accord with the actual situation and avoid subjective and arbitrary assessment. (3) Shenzhen’s industrial structure presents spatial differentiation in inner city. Unlike concentric ring model of traditional city, Shenzhen’s city center locates at the bottom of geographic space. Moreover the industrial spatial structure of Shenzhen takes clustered modern service industry as the core, spreads as fan-shaped radiation from south to north area on the whole, and has obvious path dependence characteristics. (4) The industrial spatial function of eastern division in Shenzhen is still indistinct, which means further industrial spatial structure planning and more policy support should be implemented on that region in order to develop polycentric industrial radiation mode.

Key words: urban structure, industrial function, principal component analysis, SOM clustering, Shenzhen