Weighted Oversampling Algorithm for Imbalanced Data Security Domain

To address issues such as noise, intra-class and inter- class imbalance, and data redundancy in existing methods for handling imbalanced datasets, we propose an weighted oversampling algorithm for imbalanced data security domain(WOASD). This algorithm first identifies secure, boundary, and noise samples among minority class samples based on their neighbors, removing the noise samples. It then determines the secure region of the minority class samples and calculates distance weights and density weights for boundary and secure samples. This approach enables us to ascertain the respective number of synthetic minority class samples needed for boundary and secure samples, thereby avoiding noise interference and ensuring the synthesis of secure samples. Finally, the algorithm synthesizes samples using an improved SMOTE method. Experimental results demonstrate that, compared to traditional methods, our proposed algorithm exhibits superior precision, recall, and F1 scores.