BC-Net: Early Diagnostics of Breast Cancer Using Nested Ensemble Technique of Machine Learning
Kuljeet Singh, Sourabh Shastri, Sachin Kumar, Vibhakar Mansotra · Automatic Control and Computer Sciences · 2023
Breast cancer is a divergent and prominent cancer that is responsible for the morbidity and mortality of women throughout the world. This paper aims at early detection and accurate diagnosis of this fatal disease, which is one of the most important steps in breast cancer treatment. Therefore, various nested ensemble machine learning techniques are used to help doctors determine breast cancer at an early stage. The two-layer nested ensemble model has been proposed, which encompasses stacking and voting techniques to detect benign and malignant breast cancer tumors. A total of four two-layer nested ensemble models have been proposed. S(NaiveBayes)-V(3-Meta_Learner), S(BayesNet)-V(3-Meta_Learner), S(NaiveBayes)-V(4-Meta_Learner), and S(BayesNet)-V(4-Meta_Learner) have been designed to contain base learners and meta learners. The experiments have been conducted with the k-fold cross-validation technique for model evaluation. The proposed model is capable of classifying benign and malignant breast cancer tumors with 99.50% accuracy. The aforementioned four models have been compared with previous works in terms of classification accuracy, ROC, recall, precision, TP rate, FP rate, and F1 measure. The experiments showed that the proposed two-layer nested ensemble model S(BayesNet)-V(4-Meta_Learner) performed better than the other three models mentioned supra and competed with all the previously published works. This would help the scientific community and health practitioners diagnose breast cancer with early and accurate results.