<![CDATA[Purpose: This study aims to enhance early breast cancer detection by systematically evaluating multiple machine learning (ML) algorithms and feature selection strategies. The goal is to identify the most effective combination of classifiers and feature selection methods for accurately distinguishing malignant from benign breast tumors, thereby improving diagnostic reliability and clinical decision support. Method: The Wisconsin Breast Cancer Dataset containing 699 samples described by nine diagnostic features was used. Tumor classes were encoded as 0 (malignant) and 1 (benign). The analysis was conducted in two stages. First, five ML algorithms—K-Nearest Neighbors (KNN), Logistic Regression, Support Vector Machine (SVM), Artificial Neural Network (ANN), and a hybrid SVM–ANN—were evaluated to establish baseline performance. Second, two feature selection approaches (wrapper and embedded) were applied to four ML models and the optimized hybrid classifier. The embedded approach employed Ridge-based feature selection to identify the most discriminative attributes and improve model generalization. Results: The hybrid SVM–ANN combined with Ridge Embedded feature selection achieved the best performance, with an accuracy of 97.86%, precision of 96.5%, recall of 96.5%, and an F1-score of 96%. This configuration outperformed all other algorithms and feature selection techniques, affirming the effectiveness of hybrid integration and embedded feature optimization. Novelty: The novelty lies in the integration of an SVM–ANN hybrid model with Ridge-based embedded feature selection for breast cancer classification. Unlike prior works that rely primarily on conventional filter or wrapper techniques, this approach demonstrates superior accuracy and robustness. The proposed framework provides a promising pathway for developing more reliable ML-based diagnostic tools in oncology.]]>
Copyrights © 2026
