Abstract: Bridge cables are critical load-bearing components of long-span cable-supported bridges. They are susceptible to corrosion-fatigue damage under a long-term exposure for complex environments, and their performance degradation directly affects the bridge safety and service life. Early experimental studies primarily focused on single factors, such as corrosion rate or stress ratio, often neglecting the complex interactions among multiple variables. Meanwhile, existing data-driven models exhibit limitations in prediction performance and convergence speed, along with insufficient interpretability. To address these gaps, this study establishes a hybrid machine learning SSA-LSTM-SHAP model. This approach employs the SSA-LSTM method to map the nonlinear complex relationship between fatigue performance characteristic parameters and fatigue life, while integrating SHAP technology to enhance the model convergence speed and interpretability. By integrating fatigue performance data of corroded bridge cable wires from pertinent literatures, a fatigue life database encompassing various operational parameters was constructed. The model's predictive performance was systematically validated upon this database. The study results indicate that the proposed SSA-LSTM model achieves high prediction accuracy and strong generalization capability. Furthermore, the introduction of the SHAP method addresses the "black box" nature of the SSA-LSTM model. SHAP-based interpretability reveals the specific contribution level of each input parameter to the model's predictions, providing a theoretical basis for feature engineering optimization.