The deformation mechanism of weak expansive soil slopes in open-pit mines is complex, and the spatiotemporal evolution patterns cannot be fully captured by traditional monitoring methods. Additionally, limitations are observed in existing prediction models for long-term time-series modeling and hyperparameter optimization. To address these issues, an integrated monitoring and prediction model was constructed by combining Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) technology with a Bidirectional Long Short-Term Memory network (BiLSTM) optimized by the Marine Predators Algorithm (MPA). The western slope of a large open-pit mine in western Yunnan was taken as an example. Time-series deformation information from April 2022 to March 2025 was extracted by SBAS-InSAR technology based on 76 Sentinel-1A images. The results show that the average annual vertical deformation rate of the slope in the study area ranges from -115 mm·a?1 to 85 mm·a?1, and the cumulative deformation ranges from -220 mm to 250 mm. The spatial distribution of deformation exhibits significant heterogeneity and progressive expansion characteristics, reflecting the "wet expansion and dry shrinkage" and creep deformation mechanisms of weak expansive soil under the coupled effects of wet-dry cycles and mining unloading. The deformation process shows a clear seasonal correlation with precipitation and a lag effect of 1–2 months, revealing the driving effect of water infiltration on soil strength softening. Furthermore, the hyperparameters of the BiLSTM network were globally optimized by the MPA algorithm, and the MPA-BiLSTM prediction model was constructed. Compared with the unoptimized BiLSTM model, the root mean square error and mean absolute error of the MPA-BiLSTM model are reduced by approximately 36.2% and 26.8% on average, respectively, and the coefficient of determination is generally improved. The model captures the "trend-cycle" composite deformation pattern of the slope more accurately. This study indicates that the proposed method improves the accuracy of deformation prediction, and the constructed "space-air monitoring-intelligent prediction" technical framework can provide a reliable basis for the dynamic stability evaluation and advanced disaster warning of slopes in special soil mining areas.