江苏省县域城市性发展差异的BP神经网络测定

ANALYSIS ON URBANITY DEVELOPMENT DIFFERENTIATIONS OF COUNTY AREAS IN JIANGSU PROVINCE BASED ON BP NEURAL NETWORK

本文运用人工神经网络的理论和方法,通过构建BP神经网络来评价2005年江苏省县域城市性,将52个县域城市性综合评价值分为5级。通过对频数分布特征及变异系数、加权变异系数、威廉森系数和最大与最小系数的分析,表明江苏省县域城市性空间分异显著,呈现出三个特征:①从南向北呈现梯度递减的格局;②呈正偏态分布,第三、第四级别的县市比例略大;③三大区域内部差异呈现从南向北递增的趋势。通过Spearman's rho相关分析表明,对城市性影响最大的是经济增长水平,它是提高城市性的强大动力;与城市性相关性最大的因子为X1、X4、X11和X16;因子X3和X12与城市性呈负相关。

Urbanity refers to the manifestation degree of urban nature in a certain region, which should be evaluated comprehensively by an integrated index system. Urbanity index refers to the choice of different comparison units and the determination of different evaluation criteria. Artificial Neural Network is a nonlinear system, and has good features of self-organization, self-adaption and self-learning. As one of the most widely-used networks, BP neural network takes advantages over fault tolerance, robustness and self-adaption. So this paper builds nonlinear model of BP neural network to avoid subjectivity of index weight to measure urbanity. Choosing fifty-two counties in Jiangsu province as study object, this paper first selects 16 representative indicators from the aspects of space concentration level, economic progress level, social development level and infrastructural facility construction level, and constructs index system to evaluate urbanity comprehensive value by using BP neural network theory and method according to comprehensive, representative, comparable, operational and regional principle, based on statistic data of Jiangsu province in 2005. Then, urbanity comprehensive evaluation value of fifty-two county areas are classified into five degrees, including the highest urbanity, higher urbanity, middle urbanity, lower urbanity and the lowest urbanity. Moreover, Maplnfo software is applied to make a thematic map to express the urbanity disparities of these tiff-y-two counties. Besides, the paper analyzes frequency distribution features and calculates variation coefficient, weighted variation coefficient, William coefficient and maximal and minimal coefficient, finding： 1） urbanity comprehensive evaluation value is declined from south to north; 2） its frequency distribution has positive skewness features, and a bigger proportion of counties in the third degree and fourth degree; and 3） internal differentiations of three regions increase from south to north. Finally, the paper holds that the coordination between population concentration and geographical expansion as well as synchronous development between economy and science technology should be promoted in the course of rapid urbanization. Meanwhile, space concentration level and social development level should be enhanced.