在分析灰色系统与神经网络基本原理的基础上 ,结合前人研究成果和实例分析 ,提出灰色系统用于基坑变形预测存在的一些问题 ,认为灰色系统不宜用于地下连续墙水平位移的预测 ,在其它变形预测中也要慎用。建立了基坑变形预测的神经网络模型 ,并用实例加以论证。研究表明 ,神经网络是解决基坑变形预测的有效方法 ,在地下工程中具有很好的应用前景 Based on the analysis of the basic principles of gray system and neural network, some problems existing in the prediction of foundation pit deformation are put forward based on the previous research results and case studies. It is concluded that the gray system should not be used to predict the horizontal displacement of the diaphragm wall. In other deformation prediction should also be used with caution. A neural network model of foundation pit deformation prediction is established and demonstrated with examples. The research shows that the neural network is an effective method to solve the deformation prediction of foundation pit and has a good prospect in underground engineering