Robust MRCD-PCA in Machine Learning for Breast Cancer Classification

Abstract

Breast cancer classification plays a critical role in early diagnosis and clinical decision-making. However, medical datasets are often high-dimensional and contaminated with outliers, which can degrade classification performance. While Principal Component Analysis (PCA) is commonly used for dimensionality reduction, its sensitivity to outliers limits its effectiveness in medical data analysis. To address this limitation, this study proposes a robust PCA (RPCA) approach based on the Minimum Regularized Determinant (MRCD) estimator for dimensionality reduction and named the proposed method as Robust MRCD-PCA (RMPCA). The classification using proposed RMPCA is evaluated using Support Vector Machine (SVM), Artificial Neural Network (ANN), and K-means classifiers, and compared against baseline and PCA-based models. A total of nine classification models is examined using a breast cancer dataset, with performance assessed using accuracy, precision, recall, and F1-score. Experimental results demonstrate that RMPCA achieves more consistent and reliable classification performance, particularly when combined with supervised classifiers such as SVM, outperforming both baseline and PCA-based approaches. These findings highlight the importance of robust dimensionality reduction as an effective preprocessing strategy for improving machine learning-based breast cancer classification.