人口统计数据空间化的一种方法

A Method of Spatialization of Statistical Population

人口空间分布信息在环境健康风险诊断、自然灾害损失评估和现场抽样调查比较等地理学和相关学科研究中占有重要的地位。目前随着对地观测技术和地理信息科学的飞速发展,如何精确地进行人口数据空间化成为了研究的难点和热点。针对采用传统方法解决人口空间化问题所遇到的困难和不足,设计了遗传规划(genetic programming,GP)、遗传算法(genetic algorithms,GA)和GIS相结合的方法,以GIS确定量化影响因子权重,以GP建立模型结构,以GA优化模型参数,成功建立研究区—山西省和顺县的人口数据格网分布表面。实验证明与传统建模方法(如逐步回归分析模型和重力模型)相比,所提方法建模过程更为智能化与自动化,模型结构更为灵活多样,而且数据拟合精度更高。

Mapping distribution of population has arisen as an important issue in the fields of geographical and relative researches, due to the necessity of combining with spatial data representing socio-graphic information across various spatial units, such as to evaluate the total numbers of people at environmental health risks or died in natural disasters. However, most existing solutions to this problem focus on selection and quantification of influencing factors and rarely take into account the correlation among selected factors. And much expertise is needed in modeling process to formulate the relationships between influencing factors and population data successfully. It usually not only produces information redundancy but increases the complexity of the problem. This paper explores a novel approach to transform population data from census to grid by integrating genetic programming （GP）, Genetic Algorithms （GA） and Geographic Information Systems （GIS）. A set of natural and socioeconomic factors which contribute to population distribution are identified and quantified under GIS environment. And then GP and GA are severally applied to build and optimize the population model in the hierarchical form, allowing for the computation of the relevant population data error. The experiment proves that the proposed method performs much better than stepwise regression analysis and adapted gravity model approaches. The GP/GA-based method is the first to introduce such computational intelligence techniques as GP and GA to generate gridded population maps, hence it is a methodological innovation in interpolation of population data.