Prediction of breast cancer using soft-voting based ensemble classifier

An early diagnosis of Breast Cancer (BC) is essential for reducing the mortality rate among women worldwide. Owing to the significant amount of time taken to diagnose BC, there is an urgent need for an automatic approach to diagnosing BC and its classification into benign and malignant. Although numerous models exist, there is still a need to improve their accuracy and precision. Therefore, a soft voting-based ensemble classifier is suggested in this paper to perform the categorization of whether the patient is healthy or has BC in the Breast Cancer Coimbra (BCC) dataset and benign and malignant cancer in Wisconsin Diagnostic Breast Cancer dataset (WDBC) by applying different classifiers. Owing to the strengths of classifiers, K-Nearest Neighbor (KNN), Decision Tree (DT), Support Vector Machine (SVM), and Logistic Regression (LR) are used in the ensemble classifier. The efficacy of the approach is determined in terms of performance metrics such as accuracy, recall, precision, and F1-Score. As the BCC and WDBC datasets are imbalanced, the SMOTE approach balances the datasets before applying the classifier to avoid bias. The approach achieved an accuracy of 84.61% and 98.6% for BCC and WDBC datasets. Even other parameters are better for the proposed approach than existing ones. Further, the 10-fold cross-validation achieves an accuracy of 80.45% for the BCC dataset and 98.46% for the WDBC dataset, thus demonstrating the robustness of this work.