基于BP神经网络的区域贫困空间特征研究——以武陵山连片特困区为例  

Research on Spatial Characteristics of Regional Poverty Based on BP Neural Network: A Case Study of Wuling Mountain Area

作　　者：刘一明[1,2,3] 胡卓玮[1,2,3] 赵文吉[1,2,3] 王志恒[4] 

机构地区：[1]首都师范大学资源环境与地理信息系统北京市重点实验室,北京100048 [2]首都师范大学三维信息获取与应用教育部重点实验室,北京100048 [3]首都师范大学城市环境过程与数字模拟国家重点实验室培育基地,北京100048 [4]天津城建大学地质与测绘学院,天津300384

出　　处：《地球信息科学学报》2015年第1期69-77,共9页Journal of Geo-information Science

基　　金：国家科技支撑计划项目(2012BAH33B05;2012BAH33B03;2013BAC03B04;2012BAH27B01);国家自然科学基金项目(41301468)

摘　　要：随着国家新一轮区域发展和扶贫攻坚战略的实施,连片特困地区成为新时期扶贫开发工作的主战场。本文以武陵山连片特困区县级行政区划为例,从自然和社会因素中选取主要贫困影响因子,构建评价指标体系,利用GIS和BP神经网络,模拟区域自然致贫指数、社会经济消贫指数,分析贫困的内在成因,探究贫困的空间分布特征,旨在为扶贫开发政策的制定和区域协调发展提供辅助决策。结果表明,研究区自然因素是主要的致贫原因,而社会因素在一定程度上起到了缓解作用。大部分县的自然致贫程度在中等以上,其中,铜仁、湘西地区程度较为严重,绝大多数贫困地区的社会经济水平不高,缓解贫困的能力不强;黔江地区、张家界地区的贫困程度较低,铜仁地区和湘西地区的贫困程度较高。各县的贫困状况和贫困程度存在较大差异,古丈、龙川,务川、正安,隆回、新化及道通、城步共同构成武陵山片区"大分散、小聚集"的贫困分布格局。今后的扶贫开发过程中,应充分考虑自然致贫因素,深入挖掘区域资源优势,加强区域间的交流与协作。With the implementation of new regional development and poverty alleviation strategy, contiguous destitute region shave turned into the main battle field to promote poverty alleviation and development. Selecting the contiguous destitute regions in Wuling Mountain as the study area, taking county as the study unit, this paper selects the main influence factors of poverty from common natural and social factors to build an evaluation index system. Using GIS and BP Neural Network, this paper simulates natural impoverishing index and socio-econom-ic poverty alleviation index, analyses the reason of regional poverty from the perspective of nature and society, and explores the spatial distribution characteristics of poverty in order to provide decision support for establish-ing policies for poverty alleviation and development, and achieving regional harmonious development. The re-sults show that the natural factors, such as terrain, slope, and disaster, are the main impoverishing index for the study area. The socio-economic factors, such as education, road, and medical care, could alleviate poverty to some extent. The natural poverty degree for most counties in the study area is above average, in which the Ton-gren and Xiangxi regions have relatively high level of poverty. Most destitute areas have low socio-economic lev-el, and their ability to alleviate poverty is not strong. The degree of poverty in Qianjiang and Zhangjiajie regions is lower, while in Tongren and Xiangxi regions is higher. Large differences exist between these counties’pover-ty situations. Guzhang, Longchuan, Wuichuan, Zhengan, Longhui, Xinhua, Daotong, and Chengbu together con-stitute the poverty distribution pattern of＂large dispersion, small aggregation＂in Wuling Mountain area. In the process of poverty alleviation and development, considerations should be given to the natural factors, and take advantage of local nature resources, especially the mining resources. According to the poverty type and self-de-velopment ability, different regions

关 键 词：BP神经网络 武陵山片区 贫困程度 空间分布 

分 类 号：P208[天文地球—地图制图学与地理信息工程] TP183[天文地球—测绘科学与技术]