Abstract: Predicting deformation on high slopes is crucial for ensuring the safety of open-pit mining operations. Traditionally, empirical methods and numerical simulations have been employed to predict slope displacement. However, with advancements in artificial intelligence, machine learning has emerged as an important method for predicting slope deformation in open-pit mines. Currently, a popular approach is using the backpropagation (BP) neural network to construct a time-series deformation prediction model for slopes. To enhance the BP neural network performance and prevent it from falling into local minima, a genetic algorithm (GA) is introduced to the training step of the BP neural network. This algorithm optimizes the initial weights and thresholds of the BP neural network, leading to the establishment of the time-series deformation prediction model of slopes based on the GA–BP neural network. In this paper, we collected time-series displacement data of slopes from an open-pit mine in Tibet using high-precision synthetic aperture radar (SAR) for all-weather displacement monitoring. The wavelet denoising theory was also used to eliminate interference from abnormal displacement data. The structural parameters of the BP neural network were determined using the grid search method after comparisons. For model validation, we developed two slope time-series deformation prediction models: one using the GA–BP neural network and the other using the BP neural network. We evaluated their prediction effectiveness by examining the RMSE of predicted values, the number of training operations, and model adaptability to training and validation sets. In addition, we forecasted slope deformations at five monitoring points over the next 10 h using test set data not included in the model training. The results show that the GA–BP model achieves over 10% higher accuracy than the BP model in training and validation sets. It also converges >10 times faster and adapts better to model conditions. The maximum average error of the GA–BP model on the test set is only 2.48%, with predicted displacement trends closely aligning with the monitored values. The maximum average error is 6.15%, with predictions deviating from the monitored values. In addition, the GA–BP model reduces the average prediction error at remaining monitoring points by >50% compared to the BP model. Its regression coefficients across the three datasets outperform those of the BP model, demonstrating superior generalization ability. Therefore, the GA–BP model significantly enhances the accuracy, convergence speed, and generalization ability of slope deformation predictions, offering a more reliable tool for ensuring safe production in open-pit mines.