BP神经网络在城市建成区面积预测中的应用——以江苏省为例

APPLICATION OF BP NEURAL NETWORK IN THE PREDICTION OF URBAN CONSTRUCTION LAND AREA——A CASE STUDY OF JIANGSU PROVINCE

城市建成区规模的迅速扩张是目前江苏省土地利用变化中的一个显著特点，其面积从1985年的426km。扩大到2003年的2200km^3，平均年增加98．56km^3。城市建成区规模的扩张受到社会、经济、人口等多种因素的影响，用传统方法对其进行预测比较困难。鉴于BP神经网络在非线性领域预测中的广泛应用，以江苏省为研究对象，构建一个11-3-1结构的BP神经网络预测模型，以1985～2001年和2002年的相关数据作为模型的训练和测试样本，以2003年的社会、经济数据作为网络的预测输入，对该年的建成区面积进行预测。结果表明，BP神经网络预测结果与实际面积的相对误差为3．96％，其预测精度与多元回归预测模型相比有较大改善。

The fast growth of construction land is a significant character in recent land use variation in Jiangsu province. The total area of＇its urban construction land increased from 426 km^2 in 1985 to 2 200 km^2 in 2003. The annul increase area was reached at 98. 56 km^2. Because city is a complex system and the increase of construction land area was propelled by social, economic and human factors. So it is difficult to predict the construction land area by traditional methods. In this paper, the back propagation algorithm （BP） was introduced to predict the urban construction land area in Jiangsu province. At the same time, the multi-regression analysis model was also set up to check the precision of BP neural prediction model. The results showed that the relative error between the value predicted by the BP neural network and the actual value was only 3.96%. It also showed that the BP neural network has a high precision and good effectiveness than the multi-regression analysis. The result of the research can be referred as an effective tool for the government regulations on the land use plan.