甘肃省县域城市化水平差异的人工神经网络测定

Analysis on the Difference of Urbanization Level of County Areas in Gansu Province Based on Artificial Neural Network

运用人工神经网络的理论和方法,构建BP神经网络,评价2009年甘肃省县域城市化水平,将87个县域城市化水平分为5级。对频数分布特征、变异系数、威廉森系数和最大与最小系数的分析表明,甘肃省县域城市化空间分异显著。具体表现为:呈正偏态分布,第三、四级别的县市比例较大;城市化水平发展不均衡,呈现西北—东南差异;经济区内部差异大,表现为西北高、东南低的趋势。利用Spearman’s rho相关分析得出影响城市化水平的因素及相关度。

Choosing 87 county areas in Gansu Province as study object, this paper first selects 17 representa- tive indicators from the aspects of space concentration level, economic progress level, social development level and infrastructural facility construction level, constructs index system to evaluate urbanization level by using artificial neural network theory, based on statistic data of Gansu Province in 2009. Then, urbanization level of 87 county ar- eas are classified into five degrees. Moreover, the paper analyzes frequency distribution features and calculates vari- ation coefficient, William coefficient, maximal and minimal coefficient, finding： 1 ） Its frequency distribution has positive skewness features, and a bigger proportion of counties in the third and fourth degree; 2） The urbanization level development is uneven, and it is declined from northwest to southeast ; 3 ） Internal differentiations of five eco- nomic regions declined from northwest to southeast. Finally, the Spearman＇ s rho correlation analysis indicated that the level of economic growth is the greatest impacting factor of urbanization level, which is also the powerful driving force; The proportion of non-agricultural population, per capita GDP, per capita retail sales of social consumer goods and the number per million people own a mobile phone are the most relevant factors of urbanization level, meanwhile, natural population growth rate and the number of students per million people in the school are negative- ly correlated with urbanization level.