为提高基坑变形预测精度及稳定性，首先，利用遗传算法优化BP神经网络的结构参数，再将参数优化后的BP神经网络与灰色模型结合，构建出GA-BP神经网络模型，并利用该模型实现基坑变形序列的初步预测； 其次，基于残差序列的混沌特性，再利用混沌理论进行残差优化，进一步构建考虑混沌特性优化的GA-BP神经网络模型； 最后，将SR检验引入到基坑变形趋势判断中，以检验预测结果的准确性。实例检验表明： 通过遗传算法及混沌理论的递进优化，能逐步提高预测精度，验证文章预测模型的有效性，且预测结果与SR检验结果的一致性较好，说明该预测模型的可信度高。

In order to improve the prediction accuracy of foundation pit deformation and the prediction stability, the structural parameters of BP neural network are optimized by genetic algorithm. The BP neural network is combined with grey model to establish a GABP neural network model. The preliminary prediction of the deformation sequence of the foudnation pit is realized by the model. And then the residual error is optimized by chaos theory; and the GABP neural network model considering the optimization of chaotic characteristics is further constructed based on the residual sequence. Finally, the SR test is introduced to the judgment of the deformation trend of the foundation pit to verify the accuracy of the prediction results. The case study shows that by using progressive optimization of the genetic algorithm and chaos theory, the prediction accuracy can be improved gradually, the validity of the prediction model is verified, and the consistency of the prediction results with the SR test results is better, so as to show that the reliability of the prediction model is high.