Traditional motor bearing inspection typically fails to adequately extract vibration signal features, which leads to low diagnostic accuracy. This research suggests a motor bearing fault diagnostic technique that uses a Convolutional Neural Network Combined with Support Vector Machine (CNN-SVM) and Optimized Variational Mode Decomposition (VMD) in order to increase the accuracy of motor bearing fault diagnosis. To get the best IMF classification and build the feature set, the motor bearing vibration data was modally decomposed after the Subtractive Average Optimizer (SABO) was used to tune the VMD's parameters. CNN was then employed to extract adaptive features. Lastly, CNN's fully connected layer results are categorized as SVM input. Owing to the experimental results, the SABO-VMD-CNN-SVM bearing defect diagnostic model suggested in this study has accuracy, precision, and recall rate of over 99%, which is 6.67% greater than that of the CNN-SVM model and 1.34% higher than that of the VMD-CNN-SVM version.