Breast cancer classification using machine learning requires preprocessing that can improve the representativeness of training data, particularly when class distributions are unequal. This study investigates the use of sample bootstrapping as a preprocessing step before Support Vector Machine (SVM) classification. The study uses a public breast cancer dataset containing 569 records, 30 predictor attributes, and one diagnosis label. The data are cleaned, encoded, divided into training and testing sets at a ratio of 80:20, and the training data are resampled using sampling with replacement. A linear-kernel SVM is then trained with parameter C tuning and evaluated using 10-fold cross-validation and the testing set. Based on the experimental results reported in the source study, the SVM model combined with bootstrapping achieved 98.02% accuracy, with precision, recall, and F1-score of 98.44%. The result indicates that bootstrapping is a potentially useful preprocessing strategy for improving the stability and representativeness of SVM training data. However, a numerical baseline result for SVM without bootstrapping was not documented in the source material, so the magnitude of improvement should not be interpreted beyond the reported result.
Copyrights © 2026