This paper proposes an asymmetric bagging and random subspace method (ABRS-SVM) to improve the performance of support vector machines (SVM)-based relevance feedback in content-based image retrieval (CBIR). The authors identify three main issues with SVM-based relevance feedback: 1) instability of SVM classifiers on small training sets, 2) bias in the optimal hyperplane when there are few positive feedback samples, and 3) overfitting due to high-dimensional feature spaces. To address these issues, the authors propose ABRS-SVM, which combines asymmetric bagging and random subspace methods. Asymmetric bagging is used to balance the number of positive and negative samples, while random subspace reduces overfitting by sampling features. The ABRS-SVM algorithm is evaluated against existing methods, including SVM-based relevance feedback and constrained similarity measure (CSM)-based relevance feedback. The results show that ABRS-SVM outperforms these methods in terms of precision and stability. The paper also discusses the computational complexity of the proposed algorithms and compares them with existing methods. The experiments are conducted on a large image database, and the results demonstrate that ABRS-SVM significantly improves the performance of relevance feedback in CBIR. The authors conclude that ABRS-SVM is an effective method for improving the performance of SVM-based relevance feedback in CBIR.This paper proposes an asymmetric bagging and random subspace method (ABRS-SVM) to improve the performance of support vector machines (SVM)-based relevance feedback in content-based image retrieval (CBIR). The authors identify three main issues with SVM-based relevance feedback: 1) instability of SVM classifiers on small training sets, 2) bias in the optimal hyperplane when there are few positive feedback samples, and 3) overfitting due to high-dimensional feature spaces. To address these issues, the authors propose ABRS-SVM, which combines asymmetric bagging and random subspace methods. Asymmetric bagging is used to balance the number of positive and negative samples, while random subspace reduces overfitting by sampling features. The ABRS-SVM algorithm is evaluated against existing methods, including SVM-based relevance feedback and constrained similarity measure (CSM)-based relevance feedback. The results show that ABRS-SVM outperforms these methods in terms of precision and stability. The paper also discusses the computational complexity of the proposed algorithms and compares them with existing methods. The experiments are conducted on a large image database, and the results demonstrate that ABRS-SVM significantly improves the performance of relevance feedback in CBIR. The authors conclude that ABRS-SVM is an effective method for improving the performance of SVM-based relevance feedback in CBIR.