Competitive Learning Approach to GIS Based Land Use Suitability Analysis

基于土地持续利用适宜性分析的GIS竞争学习方法研究（英文）

This paper uses the expected utility under risk hypothesis to develop a new approach to GIS modeling for land use suitability analysis with competitive learning algorithms （CLG-LUSA）. It uses Kohonen＇s Self Organ- ized Maps （SOM） and Linear Vector Quantization （LVQ） among other tools to create comprehensive ordering of high number of options. The model uses decision makers preferred locations and environmental data to construct a manifold of the decision＇s attribute space. Then, decision and uncertainty maps are derived from this manifold. An application example is provided using the selection of suitable environments for coconut development in a mu- nicipality of Cuba. CLG-LUSA model was able to provide accurate visual feedback of key aspects of the decision process, making the methodology suitable for personal or group decision making.

本文基于预期效用假设提出了可结合竞争学习算法(CLG–LUSA)的新方法,即土地利用适宜性分析的GIS模型。该模型使用了Kohonen的自组织映射法和线性矢量化法来实现多选项的综合排序。该模型还利用决策者的优选位置和环境数据,来构造一个分支决策属性空间。决策和不确定性映射来自于该分支算法。使用该模型算法的一个例子就是在古巴市选择椰子最合适的生长环境。结果表明,CLG–LUSA模型能够提供决策过程中关键环节的精确视觉反馈,从而制定最适合个人或群体决策支持方法。