To address the challenges of signal noise interference and model parameter optimization in bearing fault diagnosis, this paper proposes a comprehensive diagnostic method based on wavelet decomposition denoising, energy entropy feature extraction, and the adaptive Lévy flight improved sparrow search algorithm (ALF-SSA) optimized support vector machine (SVM). First, wavelet decomposition is used to perform multi-scale denoising on the collected bearing vibration signals, effectively suppressing noise while retaining the main features of the signal. The denoised signals are then segmented into time intervals, and the energy entropy of each segment is calculated to obtain feature vectors that reflect the local energy distribution of the signal. To overcome the limitations of the traditional sparrow search algorithm (SSA) in global search capability and local convergence, this paper introduces the adaptive Lévy flight strategy to improve the SSA, resulting in a more efficient optimization algorithm. Finally, the improved algorithm is used to optimize the key parameters of SVM, establishing a high-precision bearing fault diagnosis model. Experimental results show that the proposed method outperforms traditional methods in both diagnostic accuracy and robustness, providing effective technical support for bearing health monitoring.