Abstract: To further improve the accuracy of rolling bearing fault diagnosis, a novel method is proposed that combines TSMDE with XGBoost. First, to address issues such as the loss of informational features during the coarse-graining process of traditional multiscale entropy, a time-shift operation is introduced into the coarse-graining procedure, proposing TSMDE as a new feature extraction tool. Subsequently, the Hippo Optimization algorithm is utilized to optimize the key parameters of the XGBoost model, establishing an HO-XGBoost classifier. Finally, the proposed method was validated using the CWRU and PU bearing dataset. The simulation results demonstrate that, compared to other models, the proposed model achieved the highest diagnostic accuracies of 99.88% and 97.03%, respectively, offering a new and effective approach for rolling bearing fault diagnosis.