城市建成区规模受社会、经济、城市环境等诸多因素影响, 传统统计方法难以准确预测城 市建成区的面积。人工神经网络具有良好的非线性映射逼近性能, 在各类预测研究中得到了广泛 的应用, 尤其是BP 神经网络。主成分分析可以在有效保留数据信息前提下对数据进行降维, 它 与BP 神经网络的结合主要在数据输入端, 通过减少输入层神经元个数, 增强网络性能, 提高预 测精度。本文以北京市为例, 综合运用主成分分析和BP 神经网络方法建立预测模型, 以1986~ 2003 年数据为学习样本, 以2004 年数据为检验样本, 对2005 年北京市城市建成区面积进行模 拟预测。预测结果表明, 基于主成分分析的BP 神经网络预测结果与实际值的相对误差为2.8%, 比传统BP 神经网络预测精度提高1.8 个百分点, 网络训练收敛速度也更快, 其预测精度和效率 都有不同程度的改善。

The increase of urban built- up area is propelled by many factors of society, economy and urban environment. So it is difficult to predict the urban built- up area by traditional methods. Having good performance of nonlinear approximation, artificial neural network (ANN), especially the back propagation algorithm (BP), is applied widely in many predictions and has very satisfactory effects. Principal component analysis(PCA) can reduce the dimensions of data, while maintaining the data characteristic effectively. It is integrated with BP neural network at data input port. By decreasing the number of input neuron, it can enhance the network performance and improve the prediction. Taking Beijing for example, this article establishes a predicting model by using both PCA and BP neural networks, and makes the prediction of urban built- up area for 2005. The model’s learning samples are social, economic and environmental statistics in 1986~2003, and the testing sample is statistics in 2004. The results show that the relative error between the value predicted by the BP neural network based on PCA and the actual value is only 2.8%, and the BP neural network based on PCA has higher precision and better effectiveness than traditional BP neural network.