Breast Cancer Classification Using Support Vector Machine Method and RBF Kernel Function Based on Clinical Data, Cancer Stage, and Immunohistochemistry Results

Abstract

Background: Breast cancer is one of the leading causes of cancer-related deaths among Indonesian women. Early detection and classification of molecular subtypes are crucial for determining appropriate therapy. Accurate determination of biological subtypes of breast cancer is essential for selecting optimal treatment strategies. This research aims to build and evaluate a breast cancer subtype classification model using the SVM with an RBF kernel. The subtypes classified include Luminal A, Luminal B, HER2+, and Triple Negative Breast Cancer, utilizing a combination of patient clinical data (age, tumor size, and tumor location), cancer stage, and the expression status of hormonal receptors ER and PR. The methodological steps include data preprocessing, feature selection, model training with cross-validation, and performance evaluation using metrics such as accuracy, precision, recall, F1-score, and the ROC-AUC curve. The results showed that the majority of patients' ages were in the range of 40–60 years, with dominant tumor sizes between 1 and 3 cm. Luminal A and B subtypes were more frequently observed in patients aged ≥50 years and at early stages, whereas HER2+ and TNBC were mostly observed in patients under 50 years with advanced stages. The established baseline SVM-RBF model achieved high accuracy (91%) but performed poorly at detecting minority subtypes, such as HER2+, with a recall = 0 and an F1-score = 0, indicating model bias toward the majority class. This study demonstrates that the SVM algorithm with the RBF kernel is effective for modeling breast cancer subtype classification using clinical data, cancer stage, and immunohistochemistry results.