A method for predicting deformation during the excavation of a foundation pit in composite formation is proposed. The artificial bee colony algorithm (ABC) is introduced to optimize the back-propagation (BP) neural network with the input variables filtered. This method is applied to predict the deformation of a foundation pit project. The prediction results are verified by comparing the results with those of other neural network models. The results indicate that the depth of excavation, speed of excavation, friction angle in the soil, gravity, elastic modulus and number of internal support layers are the main factors affecting the deformation of the soil layer around the foundation pit. The ABC algorithm is capable of searching for better solutions of initial weights and thresholds. The ABC-BP model with a 6-12-2 network structure has high prediction accuracy and the best generalization ability.

提出了一种复合地层基坑开挖变形预测方法。引入人工蜂群算法 (ABC) 以优化过滤输入变量的反向传播 (BP) 神经网络。该方法用于预测基坑工程的变形。通过将结果与其他神经网络模型的结果进行比较来验证预测结果。结果表明，开挖深度、开挖速度、土体摩擦角、重力、弹性模量和内部支撑层数是影响基坑周围土层变形的主要因素。ABC算法能够寻找更好的初始权重和阈值解。