[Objective] The Beetle Antennae Search (BAS) algorithm features random oscillation and adaptive search mechanisms, demonstrating superior performance in global optimization and computational efficiency, particularly for high-dimensional complex problems. It offers unique optimization capabilities for the weight and regularization coefficients of the Hybrid Kernel Extreme Learning Machine (HKELM) model. Therefore, this study proposes utilizing BAS to optimize these two parameters to further enhance deformation prediction accuracy. [Method] Based on landslide deformation monitoring data, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) was first employed to decompose the deformation data and reduce its complexity. Subsequently, a BAS-HKELM model was constructed by combining BAS and HKELM to predict the deformation of each component. The final landslide deformation prediction was obtained by superimposing the predicted results of all components. [Results] Following CEEMDAN decomposition, the signal-to-noise ratio ranged from 43.8 to 46.8 dB, and the root mean square error ranged from 0.202 to 0.216 mm, indicating robust data decomposition capability. The BAS-HKELM prediction model achieved a mean relative error of 1.55%-1.77% and a training time of 175.09-209.33 ms. These results demonstrate both high prediction accuracy and fast convergence, validating the model’s suitability for landslide deformation prediction. [Conclusions] (1) With the progressive optimization of the CEEMDAN structure, its data decomposition capability improves gradually. Compared with db wavelet, sym wavelet, and Kalman filtering, CEEMDAN demonstrates significantly superior decomposition performance, making it suitable for processing landslide deformation data. (2) Based on data decomposition, ablation, comparative, and generalization tests consistently showed that BAS-HKELM yields superior deformation prediction performance, fully confirming its applicability to landslide deformation prediction. (3) Subsequent landslide deformation is expected to continue increasing, with accelerated growth characteristics observed at the front edge. The future deformation trend is concerning, necessitating urgent prevention and control measures.