Performance Analysis of Support Vector Machine Algorithm with Oversampling Approach in Breast Cancer Prediction

Abstract

Breast cancer is one of the most common types of cancer affecting women and is a leading cause of death in many countries. Early detection plays an important role in increasing the chances of recovery, so an accurate and reliable classification system is needed. This study aims to analyze the performance of the Support Vector Machine (SVM) algorithm in classifying breast cancer data and to evaluate the effect of applying oversampling techniques on improving model accuracy. Data imbalance between healthy and cancer-positive classes poses a challenge in the machine learning process, so the oversampling method is used to balance the distribution of training data. The test results show that the application of oversampling significantly improves model performance, with accuracy increasing from 95.91% to 98.83%, recall increasing from 89.06% to 96.88%, and F1-score from 94.21% to 98.41%, while precision remains high at 100%. This improvement shows that the combination of the SVM algorithm and the oversampling technique effectively produces a more balanced and accurate breast cancer classification system with good generalization capabilities.