An enhanced supervised learning incorporating principal component analysis for breast cancer classification and prediction
Ruxuan Liu, YiPing Yang · Communications in Statistics - Simulation and Computation · 2026
With the continuous expansion of medical data scale and the iterative development of machine learning technologies, precise prediction of breast cancer is of critical importance for clinical diagnosis and scientific treatment. Although supervised learning algorithms are widely applied, the high dimensionality and redundancy of medical data often lead to a decline in the computational efficiency of these algorithms. Principal Component Analysis (PCA) and Kernel Principal Component Analysis (KPCA), as dimensionality reduction techniques, are respectively suitable for efficient dimensionality reduction of linearly separable data and extraction of structural features from nonlinear and complex data. Based on this, this study explores the integrated application of PCA, KPCA, and supervised learning algorithms to optimize the breast cancer prediction performance. Through simulation experiments, the characteristics of supervised learning based on PCA/KPCA under limited sample conditions are verified. Furthermore, based on the Wisconsin Breast Cancer Dataset, supervised learning models are constructed, incorporating Multi-Layer Perceptron (MLP), Decision Tree (DT), Random Forest (RF), and XGBoost. Model performance is evaluated using accuracy and the Receiver Operating Characteristic (ROC) curve. The experimental results demonstrate that supervised learning models after dimensionality reduction by PCA and KPCA outperform those built with raw data across various classification metrics.