Clustering-based undersampling with random over sampling examples and support vector machine for imbalanced...

Area of Science:

• Medical Informatics
• Machine Learning
• Computational Biology

Background:

• Two-class imbalanced classification poses challenges in accurate breast cancer diagnosis.
• Existing methods may struggle with class imbalance, leading to suboptimal diagnostic performance.

Purpose of the Study:

• To propose a novel hybrid model, RK-SVM, for improved breast cancer diagnosis in imbalanced datasets.
• To enhance classification accuracy by integrating oversampling, clustering-based sample selection, and Support Vector Machine.

Main Methods:

• The proposed RK-SVM model utilizes Random Over Sampling Example (ROSE) for dataset balancing.
• K-means clustering is employed for intelligent sample selection, prioritizing samples near the class boundary.
• Support Vector Machine (SVM) is used as the core classifier to leverage the balanced and refined dataset.

Main Results:

• The RK-SVM hybrid classifier demonstrated superior performance on breast cancer datasets and other imbalanced UCI datasets.
• The method achieved higher G-mean and accuracy indices compared to existing competitive algorithms.
• Experimental results confirm the method's effectiveness in binary classification problems.

Conclusions:

• The proposed RK-SVM model offers a robust solution for imbalanced classification tasks, particularly in medical diagnosis.
• Intelligent sample selection via clustering enhances the efficiency and accuracy of oversampling techniques.
• This hybrid approach shows significant potential for improving diagnostic accuracy in breast cancer detection.