Abstract: For slope stability prediction in open-pit mines, traditional physico-mechanical analysis or numerical simulation methods suffer from complex modeling processes and high computational costs, while existing machine learning models show varying sensitivity to different data types and struggle to obtain globally optimal solutions. To address these issues, a slope stability prediction model integrating the Whale Migration Algorithm (WMA) and Light Gradient Boosting Machine (LightGBM), namely the WMA-LightGBM model, was proposed. Six primary controlling factors of slope stability—slope height, slope angle, unit weight, cohesion, internal friction angle, and pore pressure ratio—were selected as model inputs. The dual-stage collaborative optimization and adaptive migration strategy of WMA were employed to perform adaptive global optimization of LightGBM hyperparameters, enabling accurate prediction of slope stability states. Experimental results showed that the WMA-LightGBM model exhibited strong generalization ability, achieved zero missed detections of unstable slopes, and maintained a low misclassification rate for stable slopes. The model attained an accuracy of 96.3%, precision of 100%, recall of 94%, F₁-score of 0.968, and an Area Under the Curve (AUC) value of 0.98, significantly outperforming comparative models in both engineering safety and predictive accuracy. Furthermore, feature dependency analysis based on the SHAP algorithm revealed the influence patterns of input features on prediction outcomes, validating the rationality of the model's predictive logic and providing key support for its reliable engineering application in slope stability prediction scenarios.