Breast Cancer Prognosis Using Stacking Ensemble Method

摘要

This research implements an ensemble-based machine learning model, Stacking, for breast cancer prognosis. Improved breast cancer prognosis can result in faster and informed decision making by the physicians in deciding the course of treatment. The Stacking model implemented in this research combines the prediction of diverse machine learning algorithms to improve the final prediction results. Several classification studies have been performed in the field of disease prognosis focusing on the performance Accuracy. However, it is observed that if the algorithm achieves significantly high Accuracy, False Positive Rate is compromised. The resulting in an increased number of cases where the actual label is non-survival but the algorithm predicts survival. To this end, this research article focuses on improving breast cancer survival prediction by establishing a trade-off between increasing prediction accuracy and reducing the False Positive Rate. The results show that the Stacking model outperforms other methods by reducing the False Positive Rate by 5% without compromising the Accuracy.