In order to realize earlier fault detection of rolling bearings, a rolling bearing early fault detection method was proposed based on kernel entropy component analysis (KECA) and Bayesian optimization (BO) algorithm optimized support vector data description (SVDD). The time domain and frequency domain features and wavelet packet decomposition node energy features of the bearing vibration signal were extracted to form a multi-dimensional feature matrix; then, the multi-dimensional feature matrix was reduced in dimensionality by using KECA, and the effective features were extracted; finally, the characteristic indexes of the normal state of the bearing were selected to train the model, and the penalty factor and kernel width of the SVDD were determined by using the BO algorithm, and then the early fault detection model was obtained. The model was used for bearings early fault detection under different working conditions in XJTU-SY data set. The results show that KECA can effectively extract feature information and reduce the interference of redundant information. On the whole, the model can detect the occurrence of faults earlier and has better robustness and generalization ability.