基于SOM模型的中国耕地压力分类研究

根据耕地利用在经济、社会和生态方面面临的压力状况，以指标构建原则为基础，设计一套压力大小量化的指标体系，引入具有较强的聚类和容错能力的自组织特征映射(SOM)神经网络模型，在说明SOM 网络模型和算法的基础上，应用自组织特征映射网络的聚类功能，以MATLAB语言构建SOM网络模型，对我国的31省（市、自治区）耕地利用压力大小进行了分类。结果显示我国耕地压力的区域差显著且与经济地域差异有高度的一致性，表明经济发展是耕地压力的主要来源。选取大样本的神经网络训练得到的结果和现实的一致也表明，自组织特征映射模型是一种适用的耕地压力区域分类新方法。

ON THE CLUSTER OF PRESSURE LEVEL OF CULTIVATED LAND IN CHINA BASED ON SOM

Cultivated land is one of the most important natural resources in China.In order to find the way for the sustainable cultivated land-use and to overcome the default of regular cultivated land-use pressure,a logging theological identification technology is introduced and the principle of SOM(self-organizing mapping) network is summarized in this paper.Then a set of cultivated land-use pressure evaluation indicators based on the indicator selection principles are set up according to the cultivated land current condition in China in terms of economy,society and ecology.A program is written,and MATLAB 6.1 software is employed to measure the cluster of regional cultivated land use pressure.The run results of SOM show that 31 provinces(cities) or autonomous regions are classified into various categories on the condition of different steps.The regional cultivated land pressure level disparities are consistent with the economic regional difference.Cultivated land pressure status estimated on SOM can explore distinctively the reason and the results of cultivated land-use pressure changes,which will help administrators to adopt suitable land-use policy and management measures to alleviate cultivated land use pressure and improve cultivated land quality.The evaluation results also indicate that the application of SOM neural network in assessing regional differentiations of cultivated land use pressure level without assuming parametric relationship is convenient,precise and feasible,which can be an alternative approach of assessing regional differentiations of cultivated land use pressure level.