非线性多维数据可视化分类预测方法

Visual classification prediction method for nonlinear and multi-dimensional data

地理信息分类的传统线性算法具有正向直接判定的快速优势,但局限于对已知数据进行线性的判别划分,而非线性未知信息的分类预测同样是GIS技术的重要内容。人工神经网络算法为一些非线性知识的发现提供了可能。本文在通用的GIS格式数据基础上,采用L-M算法进行分类,通过分类结果来预测未知信息。开发出可视化的GIS数据神经网络分类预测软件模块。并以美国各镇人口为样例数据进行测试,分类预测结果显示该算法具有可行性及系统具有实用性。

Traditional linear classification algorithms of GIS have the advantage of determining in positive direction quickly, but have the limitation in determining the given data linearly. However, the classification prediction for the unknown non-linear information is the important part of GIS. Artificial neural network algorithm makes it possible to discover some non-linear knowledge. Based on the common format of GIS data, the paper adopted L-M algorithm to do classification and to classify and predict the unknown information further through the classification results. A neural network classification prediction visualization module based on GIS was constructed, which was tested using the U. S. towns and population data. It is proved that the L-M algorithm is feasible and the module is of practicality.