城市土地利用变化 ,具有非线性特征 ,一般的数据挖掘方法基本上失效。本文研究了利用马尔可夫链和神经网络两种方法 ,基于地理信息系统、遥感图像、电子地图 ,预测了上海市中心城区、 2 0 0 2年和 2 0 0 5年的土地利用总量和土地利用类型结构的变化 ,从而研究了城市土地利用状况演变预测的地学数据挖掘技术。

This paper focuses on GIScience-techniques with the involvement of two technques of data mining,the Markov Chain method and Artificial Neural Network(ANN)method.The results were used to the strategic planning and development of the city of Shanghai.In order to estimate state transtation matrix,firstly the city was divided into 5 zones based on GIS,remote sensing images and electronic maps,and then the total amount of land use in Shanghai in 2002 and 2005 was estimated by Markov Chain method. Secondly,changes in land use types in core areas of Shanghai were forecasted with ANN method which shows the ANN Model 2 (Fig.3)is better than Model 1(Fig.1).This indicates that each one has its strong point,for the ANN method can be better used to forecast directional aspect of land use,and the Markov Chain method can be better used to forecast changes of land use within the zones.Here Markov Chain method was found that is wreath to take the inside variety, ANN considered fit the land that make use in class direction is good.It is further found out that merely for the converted land use types,the standard ANN-BP method gives greater error.This can be explained as the transform process of land use types of Shanghai city is directional,a general ANN-BP arithmetic gives each land type conversion with the same weight function.Although it is self-contradictory in directionality of the transformation,the modified model can still overcome this difficulty.