Abstract: In this paper, an improved northern goshawk optimization algorithm (IMNGO) is proposed to optimize the variational mode decomposition (VMD) and support vector machine (SVM) for small-sample bearing fault identification. This algorithm can effectively solve the problems of insufficient information extraction and few available fault samples when the rolling bearing fails. The experimental data is collected through the cloud platform, and the VMD parameters are optimized using the IMNGO algorithm to find the best intrinsic mode component (IMF), construct the eigenvector energy spectrum and the principal component contribution map, and screen the best IMF component. Finally, the extracted feature information is imported into the SVM optimized by IMNGO for small sample detection and recognition of bearings. After IMNGO optimization, the recognition accuracy rate under single working conditions reached 99.20%, and the recognition accuracy rate under complex working conditions reached 94.45%. Under the small sample data, the method proposed in this paper has greatly improved the recognition accuracy compared with the traditional detection method.