Remaining useful life (RUL) prediction is a core task in prognostics and health management. While Transformers excel at modeling long-range temporal dependencies, their performance is highly sensitive to hyperparameters, and improper splitting of sliding-window samples can introduce data leakage. We propose a Sparrow Search Algorithm (SSA)-optimized Transformer for CMAPSS RUL prediction, adopting an engine-wise split for leakage-aware model selection and using validation RMSE as the fitness function to guide SSA-based hyperparameter optimization. On the FD001 test set, the model achieves RMSE $13.79$, MAE $10.00$, $R^2=0.88$, and a NASA score of $356.26$. The prediction curves and residual d... More >