At present, the foundation pit engineering generally only cares whether the ground settlement value exceeds the monitoring and early warning value, and it lacks an effective method for short-term and real-time prediction of foundation pit surface settlement, which reduces the safety of the foundation pit. Using artificial bee colony algorithm to optimize the combination model of BP neural network can reasonably predict the settlement of the foundation pit surface. First, the gray correlation theory was combined to filter the input variables to improve the efficiency of the network structure. Then, the artificial bee colony algorithm was used to optimize the initial value of the BP neural network to realize the prediction of the cumulative maximum value and location of the surface settlement. Finally, ABC-BP model was compared with other common neural network prediction models to verify the validity of the model. It can be seen from the prediction and comparison results that the average relative error of ABC-BP model training and prediction results is 3.27%, and the root mean square error is 3.87, verifying that the model is effective.