With the development of intelligent operation and maintenance of industrial equipment, bearing fault diagnosis has emerged as a crucial factor in ensuring the reliability of mechanical systems. This paper designs a comprehensive fault diagnosis model for source - domain data based on the extracted fault features. First, a high - quality source - domain dataset is constructed through data preprocessing, feature screening, and file grouping. Subsequently, a radial basis function kernel support vector machine with class weights (RBF - SVM) is employed as the diagnosis model, and its performance is optimized through hyperparameter tuning. The experimental results indicate that the model achieves an overall accuracy of 93.18% and a macro - average F1 score of 0.9278 on the test set, and demonstrates high generalization ability and mechanism consistency in multi - dimensional analysis. The work in this paper validates the effectiveness of feature extraction and lays a foundation for subsequent transfer learning tasks.