Improving Minority Class Detection in Breast Cancer Dataset using SMOTE with SVM and Adaboost

Abstract

Imbalanced data often hinders the ability of classification models to accurately detect minority classes, leading to biased predictions toward the majority class. This study aims to enhance the detection performance of minority classes in Breast Cancer data by integrating the Synthetic Minority Over-sampling Technique (SMOTE) with two classification algorithms: Support Vector Machine (SVM) and Adaptive Boosting (AdaBoost). The Breast Cancer dataset from the UCI Machine Learning Repository is used due to its imbalanced class distribution. The proposed approach applies SMOTE at the preprocessing stage to generate a more balanced dataset before model training. Both SVM and AdaBoost are then evaluated on the original and SMOTE-balanced data using three performance metrics: accuracy, sensitivity, and F1 score. Experimental results demonstrate that the application of SMOTE significantly improves the classifiers’ ability to recognize minority classes. Furthermore, AdaBoost achieves superior performance compared to SVM in terms of both overall accuracy and minority class sensitivity.